MemoryLane v0.26.1

Internal hardening plus a smoother enterprise activation flow when the backend already has consent on file.

What's Changed

• Enterprise: skip the in-app consent dialog when the backend reports already_approved; the client binds via external consent and proceeds straight to polling.
• Tightened how custom vendor baseURL values and backend-returned URLs are validated.
• Stricter validation of the database export directory (must be absolute and inside an allowed location).
• Customer checkout and subscription portal now open via short-lived signed URLs minted from the backend.
• Release workflow resolves the input ref to an immutable commit SHA once and reuses it across every job; all GitHub Actions are pinned to commit SHAs and tracked by Dependabot.

Known Issues & Limitations

• Vertex managed-mode bearer tokens aren't refreshed in-flight — long-running operations that outlive the token TTL may see 401s until the next refresh cycle (DEU-84).
• Windows OCR still depends on native OCR component availability.
• Intel macOS is not yet officially supported.

Installation

• macOS customer (Apple Silicon): install from the GitHub release page (auto-update enabled).
• macOS enterprise (Apple Silicon): MemoryLane Enterprise-arm64-mac.pkg — delivered privately per customer.
• Windows customer: MemoryLane-Setup.exe
• Windows enterprise: MemoryLane Enterprise-Setup.msi — delivered privately per customer.

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.26.0...v0.26.1

MemoryLane v0.26.0

Reworks the LLM client layer behind a single InferenceProvider, adds Google Vertex AI for enterprise, remembers model selections per vendor, and bakes per-customer backend URLs into enterprise builds at build time.

What's Changed

• Google Vertex AI support (enterprise) (#130): the enterprise edition can now route LLM calls through Vertex via the license-issued inference config.
• Per-vendor model memory (#130): swapping vendors no longer resets your model picks — each vendor remembers its own snapshot, video, and pattern-detection model.
• Custom endpoint extended to pattern detection and user-context (#130): an OpenAI-compatible local endpoint (LM Studio / Ollama) now drives all three LLM features, not just semantic summaries.
• Unified LLM client provisioning (#130): internal refactor onto Vercel AI SDK via a shared InferenceProvider. Removes the legacy api-key-manager / custom-endpoint-manager split in favour of a single vendor-credentials.json; existing configs migrate automatically on first launch.
• summary_model activity column (#130): each activity now records which model produced its summary.
• Electron 40.1.0 → 40.9.2 (#131).

Fixes

• User-context model now comes from per-vendor settings instead of the legacy global slot (#133).

Known Issues & Limitations

• Vertex managed-mode bearer tokens aren't refreshed in-flight — long-running operations that outlive the token TTL may see 401s until the next refresh cycle (DEU-84).
• Windows OCR still depends on native OCR component availability.
• Intel macOS is not yet officially supported.

Installation

• macOS customer (Apple Silicon): install from the GitHub release page (auto-update enabled).
• macOS enterprise (Apple Silicon): MemoryLane Enterprise-arm64-mac.pkg — delivered privately per customer.
• Windows customer: MemoryLane-Setup.exe
• Windows enterprise: MemoryLane Enterprise-Setup.msi — delivered privately per customer.

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.25.0...v0.26.0

MemoryLane v0.25.0

Reshapes the enterprise activation flow around a single user-facing activation code and a Bearer-auth backend contract.

What's Changed

• Single activation code for enterprise (#128): users now paste one tt_<token>.<email> activation code instead of separate fields. The device parses it locally, sends tenant_token and email to the backend, and uses Bearer auth on /license/* and /device/upload endpoints. The consent document is fetched up-front via a descriptor, hash-verified, and pinned to the configured backend origin to prevent off-host redirection. 401 on /status and /key is now treated as inactive rather than an error.
• Skip semantic client rebuild on no-op key updates (#127): updateApiKey no longer rebuilds the embedding client when the key hasn't changed.

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Intel macOS is not yet officially supported

Installation

• macOS customer (Apple Silicon): install from the latest GitHub release or via the project install script
• macOS enterprise (Apple Silicon): MemoryLane Enterprise-arm64-mac.pkg — delivered privately
• Windows customer: MemoryLane-Setup.exe
• Windows enterprise: MemoryLane Enterprise-Setup.msi — delivered privately

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.24.0...v0.25.0

MemoryLane v0.24.0

Adds an enterprise consent step before a device is bound to an activation key.

What's Changed

• Enterprise consent step on activation (#126): the activate endpoint is now a probe — when the server returns a consent document, the app renders it (PDF) inline, gates Accept on a checkbox, and only binds the device after the user accepts. Decline returns to key entry; a 15-minute timeout protects against abandoned decisions. Probe responses are restricted to PDF, and the downloaded document is verified against the sha256 from the probe before it is shown.

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Intel macOS is not yet officially supported

Installation

• macOS customer (Apple Silicon): install from the latest GitHub release or via the project install script
• macOS enterprise (Apple Silicon): MemoryLane Enterprise-arm64-mac.pkg — delivered privately
• Windows customer: MemoryLane-Setup.exe
• Windows enterprise: MemoryLane Enterprise-Setup.msi — delivered privately

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.23.8...v0.24.0

MemoryLane v0.23.8

Fix the MCP server crashing on start when launched by Claude Desktop.

What's Changed

• MCP server no longer crashes with Cannot find module 'electron': a top-level import { app } from 'electron' in edition.ts was code-split into a shared chunk that the MCP entry transitively required. Under ELECTRON_RUN_AS_NODE=1 the electron module is unresolvable, so the process died before main() ran. Electron APIs are now read through a lazy, shared getElectronAppOrNull() helper, and a compiler-API-based guard test fails if any file in the MCP import graph reintroduces a top-level electron import.

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Intel macOS is not yet officially supported

Installation

• macOS customer (Apple Silicon): install from the latest GitHub release or via the project install script
• macOS enterprise (Apple Silicon): MemoryLane Enterprise-arm64-mac.pkg — delivered privately
• Windows customer: MemoryLane-Setup.exe
• Windows enterprise: MemoryLane Enterprise-Setup.msi — delivered privately

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.23.7...v0.23.8

MemoryLane v0.23.7

Edition-aware default DB path for the MCP server.

What's Changed

• Enterprise MCP uses the enterprise DB by default: on the enterprise build, productName = "MemoryLane Enterprise" shifts Electron's userData directory, but the MCP server's default DB path was hardcoded to the customer location. Enterprise users therefore saw an empty DB through MCP with no obvious workaround. The default path is now edition-aware across macOS, Windows, and Linux (dev and packaged).
• New get_db_path MCP tool: returns { path, source, edition, defaultForEdition } so Claude can verify which DB is being read and decide whether to call set_db_path / reset_db_path.
• Tray tooltip shows the app version: hovering the tray icon now includes v<version>, and the tray menu has a disabled version entry.

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Intel macOS is not yet officially supported

Installation

• macOS customer (Apple Silicon): install from the latest GitHub release or via the project install script
• macOS enterprise (Apple Silicon): MemoryLane Enterprise-arm64-mac.pkg — delivered privately
• Windows customer: MemoryLane-Setup.exe
• Windows enterprise: MemoryLane Enterprise-Setup.msi — delivered privately

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.23.6...v0.23.7

MemoryLane v0.23.6

Claude Desktop integration fix for Windows MSIX installs.

What's Changed

• Claude Desktop integration on Windows MSIX: fresh MSIX-packaged Claude Desktop reads its config from a virtualized sandbox under %LOCALAPPDATA%\Packages\Claude_*\LocalCache\Roaming\Claude\ instead of %APPDATA%\Claude\. MemoryLane now discovers every Claude_* package and dual-writes, so the tray "Connect to Claude Desktop" button actually shows up in Claude's server list. Users upgraded from an earlier MemoryLane will see a "Reconnect" prompt that backfills the MSIX path.

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Intel macOS is not yet officially supported

Installation

• macOS customer (Apple Silicon): install from the latest GitHub release or via the project install script
• macOS enterprise (Apple Silicon): MemoryLane Enterprise-arm64-mac.pkg — delivered privately
• Windows customer: MemoryLane-Setup.exe
• Windows enterprise: MemoryLane Enterprise-Setup.msi — delivered privately

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.23.5...v0.23.6

MemoryLane v0.23.5

App is now the MCP entrypoint, with multi-DB support.

What's Changed

• No more Node prerequisite for integrations: one-click setup for Claude Desktop, Claude Code, and Cursor now runs the MemoryLane app directly under ELECTRON_RUN_AS_NODE=1 instead of shelling out to npx. Integrations keep working on machines without Node installed.
• Reconnect flow for stale entries: if the app was moved or upgraded, the Integrations panel surfaces a "Reconnect" button instead of silently rewriting config. User-added --db-path args are preserved when reconnecting.
• Multi-DB via set_db_path / reset_db_path: the MCP server can now be pointed at a different MemoryLane database at runtime. The recorder always writes to the default DB — set_db_path only affects what the MCP server reads.

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Intel macOS is not yet officially supported

Installation

• macOS customer (Apple Silicon): install from the latest GitHub release or via the project install script
• macOS enterprise (Apple Silicon): MemoryLane Enterprise-arm64-mac.pkg — delivered privately
• Windows customer: MemoryLane-Setup.exe
• Windows enterprise: MemoryLane Enterprise-Setup.msi — delivered privately

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.23.4...v0.23.5

MemoryLane v0.23.4

macOS enterprise edition and exclusions UI polish.

What's Changed

• macOS enterprise (.pkg): the enterprise edition now ships for macOS as a signed + notarized .pkg installer alongside the Windows MSI. Same runtime as Windows enterprise — license activation, no auto-update, optional DB upload sync.
• Found banner dismissal persists: collapsing the "Found (N)" block now sticks across window reopens within a session.
• Self-filter works in packaged builds: MemoryLane no longer shows up in its own Found list for customer/enterprise bundles.

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Intel macOS is not yet officially supported

Installation

• macOS customer (Apple Silicon): install from the latest GitHub release or via the project install script
• macOS enterprise (Apple Silicon): MemoryLane Enterprise-arm64-mac.pkg — delivered privately
• Windows customer: MemoryLane-Setup.exe
• Windows enterprise: MemoryLane Enterprise-Setup.msi — delivered privately

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.23.1...v0.23.4

MemoryLane v0.23.3

macOS enterprise edition and exclusions UI polish.

What's Changed

• macOS enterprise (.pkg): the enterprise edition now ships for macOS as a signed + notarized .pkg installer alongside the Windows MSI. Same runtime as Windows enterprise — license activation, no auto-update, optional DB upload sync.
• Found banner dismissal persists: collapsing the "Found (N)" block now sticks across window reopens within a session.
• Self-filter works in packaged builds: MemoryLane no longer shows up in its own Found list for customer/enterprise bundles.

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Intel macOS is not yet officially supported

Installation

• macOS customer (Apple Silicon): install from the latest GitHub release or via the project install script
• macOS enterprise (Apple Silicon): MemoryLane Enterprise-arm64-mac.pkg — delivered privately
• Windows customer: MemoryLane-Setup.exe
• Windows enterprise: MemoryLane Enterprise-Setup.msi — delivered privately

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.23.1...v0.23.3

MemoryLane v0.23.1

Privacy and Windows stability fixes.

What's Changed

• Auto-fill is now opt-in: discovered apps/sites appear in a "Found (N)" block with toggles off — you pick what to block, or click Add all. Previously items were auto-added to exclusions (#121).
• Windows MSI upgrades unbroken: native helper processes are now killed on quit so the installer can replace them during upgrade (#120).
• Per-tab dismiss: collapsing the Found block for Apps no longer collapses it for Websites (and vice versa).

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Intel macOS is not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download the latest GitHub release and use either MemoryLane-Setup.exe or MemoryLane-Setup.msi

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.23.0...v0.23.1

MemoryLane v0.23.0

Faster exclusion setup and better Windows app matching.

What's Changed

• Auto-fill exclusions from activity: new "Auto-fill from activity" flow watches app/URL changes for 2 minutes (screen capture suppressed while running) and pre-populates the exclusion lists. A "Just added" pinned section lets you un-toggle anything captured by mistake.
• Search-first websites picker: default view shows currently-blocked domains; the search input doubles as add-custom. Matches come from the union of excluded and seen domains.
• Windows apps picker parity: Start Menu scanner now resolves .lnk targets and uses the target exe stem as the match token, so toggles actually take effect at runtime. Noise entries like .msc snap-ins and release notes are dropped (#119).
• Privacy hardening: observation controller edge cases, unified host filter, newtab filtered at both entry points, cache TTL on installed-apps list.
• App-watcher fans out to multiple subscribers so observation and interaction-monitor share one native watcher.

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Intel macOS is not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download the latest GitHub release and use either MemoryLane-Setup.exe or MemoryLane-Setup.msi

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.22.1...v0.23.0

MemoryLane v0.22.1

Remote sync is now opt-in.

What's Changed

• Enterprise remote sync defaults to Off on fresh installs. The Summary/Detailed toggle becomes a three-state switch: Off / Summary / Detailed. When Off, both the 24h periodic upload and the manual "Sync to Remote" button are suppressed. Existing users keep their prior selection
• Strict gating: the sync gate is re-evaluated on every tick and before every manual upload

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Intel macOS is not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download the latest GitHub release and use either MemoryLane-Setup.exe or MemoryLane-Setup.msi

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.22.0...v0.22.1

MemoryLane v0.22.0

Upload privacy controls and Slack removal.

What's Changed

• Enterprise uploads now support a Summary/Detailed toggle — Summary (default) strips OCR text and FTS index; Detailed preserves them. Both modes strip personal context and pattern detection runs
• Removed Slack integration

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Intel macOS is not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download the latest GitHub release and use either MemoryLane-Setup.exe or MemoryLane-Setup.msi

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.21.0...v0.22.0

MemoryLane v0.21.0

Enterprise privacy hardening and backend URL improvements.

What's Changed

• Database uploads now strip sensitive data (OCR text, FTS indexes) before sending to the enterprise backend, keeping only activities, patterns, sightings, and vector embeddings
• Enterprise backend URL uses /api prefix and is configurable via MEMORYLANE_BACKEND_URL environment variable

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Intel macOS is not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download the latest GitHub release and use either MemoryLane-Setup.exe or MemoryLane-Setup.msi

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.20.3...v0.21.0

MemoryLane v0.20.3

Fixes enterprise database upload failing with a CSRF error.

What's Changed

• Fixed the database upload endpoint path to match the server's routing convention, resolving CSRF errors on upload
• Enterprise edition now skips the MCP connect step during onboarding

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Intel macOS is not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download the latest GitHub release and use either MemoryLane-Setup.exe or MemoryLane-Setup.msi

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.20.2...v0.20.3

MemoryLane v0.20.2

Fixes a startup crash on Windows caused by onnxruntime DLLs not being found.

What's Changed

• Moved the onnxruntime DLL PATH fix into a dedicated side-effect module (onnxruntime-path-fix.ts) imported before all other modules, so it runs before the static import chain (runtime → embedding → @huggingface/transformers) triggers require('onnxruntime-node')

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Intel macOS is not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download the latest GitHub release and use either MemoryLane-Setup.exe or MemoryLane-Setup.msi

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.20.1...v0.20.2

MemoryLane v0.20.1

Fixes a startup crash on Windows caused by missing onnxruntime DLLs.

What's Changed

• Fixed Windows startup crash: onnxruntime DLL directory is now added to PATH before module load, so the Windows DLL loader can find onnxruntime.dll and DirectML.dll in the asar.unpacked path

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Intel macOS is not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download the latest GitHub release and use either MemoryLane-Setup.exe or MemoryLane-Setup.msi

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.20.0...v0.20.1

MemoryLane v0.20.0

Enterprise edition: automatic and manual database sync to remote.

What's Changed

• Added periodic database upload sync for enterprise edition — activated devices upload a backup to the enterprise backend every 24 hours
• Added a "Sync to Remote" button in Advanced Settings for enterprise users to trigger an upload manually
• Renamed ENTERPRISE_LICENSE_CONFIG to ENTERPRISE_BACKEND_CONFIG to reflect its broader scope

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Intel macOS is not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download the latest GitHub release and use either MemoryLane-Setup.exe or MemoryLane-Setup.msi

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.19.1...v0.20.0

MemoryLane v0.19.1

Patch release: fixes a stuck Claude / Cursor MCP connection after upgrading from a pre-v0.18 build, and improves the CLI's diagnostics.

What's Changed

• Fixed startup MCP migration: stale memorylane entries pointing at the deleted in-asar mcp-entry.js are now rewritten to the standalone CLI invocation, even when the original env block is missing
• Added a permission prompt for system notifications during screen recording (#DEU-15)
• CLI now surfaces an actionable hint when better-sqlite3 is missing its native binary
• CLI README documents the Node LTS requirement and bindings troubleshooting steps

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Intel macOS is not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download the latest GitHub release and use either MemoryLane-Setup.exe or MemoryLane-Setup.msi

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.19.0...v0.19.1

MemoryLane v0.19.0

New stepped onboarding flow that guides users from MCP setup through first pattern detection.

What's Changed

• Added stepped onboarding: Connect (pick MCP providers) -> Capture (progress bar) -> Dashboard
• Added openExternal IPC for opening links in the default browser (https-only)
• Added estimated hours per week to pattern cards
• Increased minimum activities before pattern detection from 50 to 200
• Renamed "Claude Desktop" to "Claude Cowork" in integrations
• Simplified StatusLine to show capture state with animated indicator

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Intel macOS is not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download the latest GitHub release and use either MemoryLane-Setup.exe or MemoryLane-Setup.msi

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.18.0...v0.19.0

MemoryLane v0.18.0

Minor release: enterprise edition support, MCP CLI cleanup, and plugin workflow updates.

What's Changed

• Added edition-aware customer and enterprise build flow with activation support
• Moved MCP into the standalone CLI package and added public HTTP mode with ngrok/token auth
• Converted plugin commands into skills and added a safer deploy-plugin release workflow
• Fixed plugin marketplace validation by normalizing the plugin name

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Intel macOS is not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download the latest GitHub release and use either MemoryLane-Setup.exe or MemoryLane-Setup.msi

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.17.4...v0.18.0

MemoryLane v0.17.4

Patch release: updated Explorer plan price to $50/mo.

What's Changed

• Updated Explorer plan price from $30/mo to $50/mo

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Linux and Intel macOS are not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download the latest GitHub release and use either MemoryLane-Setup.exe or MemoryLane-Setup.msi

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.17.3...v0.17.4

MemoryLane v0.17.3

Patch release: renamed the Power User plan to Explorer.

What's Changed

• Renamed subscription plan from "Power User" to "Explorer"

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Linux and Intel macOS are not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download the latest GitHub release and use either MemoryLane-Setup.exe or MemoryLane-Setup.msi

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.17.2...v0.17.3

MemoryLane v0.17.2

Patch release: managed subscription keys refresh in the background so entitlement stays aligned with the backend.

What's Changed

• Added a daily background refresh for provisioned API keys so subscription changes apply without restarting the app

Features

• Managed-key subscription state stays in sync via periodic re-fetch (24-hour interval)

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Linux and Intel macOS are not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download the latest GitHub release and use either MemoryLane-Setup.exe or MemoryLane-Setup.msi

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.17.1...v0.17.2

MemoryLane v0.17.1

Patch release focused on safer Windows updates, clearer pattern language, and docs cleanup.

What's Changed

• Blocked Windows auto-update installs while the MCP host is running to avoid interrupted sessions
• Updated pattern prompts and cards to prefer clearer first-person descriptions
• Refreshed the README, privacy link, and CLI/plugin documentation

Features

• Windows update handling is safer when MemoryLane is active through MCP
• Pattern recommendations use more natural descriptions across prompts and UI cards
• The MemoryLane plugin now includes the process-flowchart command and skill

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Linux and Intel macOS are not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download the latest GitHub release and use either MemoryLane-Setup.exe or MemoryLane-Setup.msi

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.17.0...v0.17.1

MemoryLane v0.17.0

Minor release focused on pattern workflow UX and MCP prompt cleanup.

What's Changed

• Added an expandable pattern feedback nudge card for quicker review and response actions
• Improved pattern list behavior and persisted activity updates for smoother in-app feedback loops
• Removed MCP prompt payloads from server responses to keep context output cleaner

Features

• Task recommendation cards now support a compact/expanded feedback flow with thumbs up/down handling
• MCP context responses are leaner and better suited for downstream assistants

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Linux and Intel macOS are not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download the latest GitHub release and use either MemoryLane-Setup.exe or MemoryLane-Setup.msi

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.16.0...v0.17.0

MemoryLane v0.16.0

Minor release that keeps the core feature set stable while rolling back recent UI and updater behavior changes.

What's Changed

• Reverted the pattern review interface back to the flat list with thumbs up/down actions
• Reverted updater recheck-before-install behavior to the previous install flow
• Reverted release workflow Discord notification changes

Features

• Same feature set as v0.15.6

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Linux and Intel macOS are not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download the latest GitHub release and use either MemoryLane-Setup.exe or MemoryLane-Setup.msi

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.15.6...v0.16.0

MemoryLane v0.15.4

Patch release focused on approved automation visibility.

What's Changed

• Approved patterns stay visible - Approving a low-sighting automation no longer makes it disappear from the reviewed list
• Sightings filter stays accessible - The filter bar now remains visible whenever approved automations exist

Features

• Reviewed automation cards remain visible immediately after approval
• Sightings filter stays available for all approved automation opportunities

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Linux and Intel macOS are not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download the latest GitHub release and use either MemoryLane-Setup.exe or MemoryLane-Setup.msi

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.15.3...v0.15.4

MemoryLane v0.15.3

Patch release focused on workflow pattern review UX and visibility.

What's Changed

• Stacked review cards for unreviewed patterns - Pattern suggestions now use a clearer card stack to speed up triage
• Completed state support in pattern cards - Reviewed workflows now show explicit completion status
• Cleaner reviewed-pattern views - Sighting filters are hidden when there are no reviewed patterns to filter

Features

• Stacked review UI for unreviewed pattern recommendations
• Completed-state presentation for pattern cards
• Less noisy filtering controls in empty reviewed-pattern states

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Linux and Intel macOS are not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download the latest GitHub release and use either MemoryLane-Setup.exe or MemoryLane-Setup.msi

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.15.2...v0.15.3

MemoryLane v0.15.2

Patch release focused on capture reliability, activity timing correctness, and pattern consistency.

What's Changed

• More reliable macOS capture - Screenshot sidecars now restart more safely and avoid unbounded memory growth when frame writing falls behind
• More accurate activity boundaries - Manual capture start now schedules background analyzers correctly, typing duration calculation is fixed, and single-item sampling behaves correctly
• Pattern consistency fixes - Pattern counts, search results, and stored pattern state stay aligned more reliably
• Release pipeline polish - Release notifications are now sent to Discord with tighter workflow guards

Features

• Safer macOS screenshot capture under crash and backpressure conditions
• Corrected activity timing and capture-start behavior in the background pipeline
• Better pattern normalization so detection, counts, and search stay in sync
• Release workflow notifications for shipped builds

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Linux and Intel macOS are not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download the latest GitHub release and use either MemoryLane-Setup.exe or MemoryLane-Setup.msi

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.15.1...v0.15.2

MemoryLane v0.15.1

Patch release focused on pattern detection controls, a stronger detection pipeline, and integration/status polish.

What's Changed

• Pattern detection controls - Pattern detection can now be turned on or off from settings, with a clearer main window call to action when it is disabled
• Stronger detection pipeline - Pattern detection now uses a two-phase agentic flow with tool access, refreshes on window focus, and prunes stale patterns more reliably (#87)
• Integration and status polish - MCP status checks are now real-time, browser-native alerts were replaced with app UI, and startup/settings behavior is cleaner
• Patterns and model UI cleanup - Pattern cards surface duration estimates more clearly, while model selectors and integration controls stay out of the way unless needed

Features

• Toggle pattern detection without leaving the app flow
• Improved agentic pattern detection with stale-pattern pruning and better refresh timing
• Live MCP status checks and fewer disruptive dialogs in setup flows
• Clearer pattern cards, integrations UI, and model controls in settings

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Linux and Intel macOS are not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download the latest GitHub release and use either MemoryLane-Setup.exe or MemoryLane-Setup.msi

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.15.0...v0.15.1

MemoryLane v0.15.0

Minor release focused on the redesigned settings experience, auto-generated user context, live pattern feedback, and clearer model health/status signals.

What's Changed

• Settings overhaul - Advanced settings are consolidated into clearer sections for models, privacy, connections, and capture controls, with model edits autosaving correctly
• User context builder - MemoryLane now builds a reusable user context profile from your activity history to improve semantic understanding
• Pattern feedback loop - Pattern cards are wired to live database data with persistent thumbs up/down feedback and rejection tracking
• Model health and status clarity - The app now surfaces LLM health checks, a simpler status line, and better custom endpoint prefills and validation
• MCP and plan polish - Claude setup guidance is easier to collapse when not needed, plan defaults were streamlined, and the plugin integration docs were refreshed

Features

• Reworked advanced settings UI with clearer grouped sections and better autosave behavior
• Auto-generated user context can now be derived from activity data and reused by semantic features
• Pattern suggestions now use live stored data and keep approve/reject feedback across sessions
• LLM health checks and status indicators make provider/configuration issues easier to spot
• Claude and plugin setup flows are cleaner for users who do not need every integration path

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Linux and Intel macOS are not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download the latest GitHub release and use either MemoryLane-Setup.exe or MemoryLane-Setup.msi

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.14.3...v0.15.0

MemoryLane v0.14.3

Patch release focused on custom semantic endpoints, better privacy controls, improved release packaging, and periodic raw database export.

What's Changed

• Custom semantic endpoints - OpenAI-compatible providers now work directly for semantic processing, with better provider error handling and cached video fallback decisions
• Privacy rule matching - App names, window titles, and URL patterns now match substrings by default, making exclusions easier to configure
• Periodic raw DB export - Added a settings-controlled raw SQLite export folder for regular backup snapshots alongside the existing manual ZIP export
• CLI and plugin MCP flow - CLI now supports MCP server mode so the plugin can work without the desktop app running
• Windows release packaging - Release pipeline now publishes both MemoryLane-Setup.exe and MemoryLane-Setup.msi with stable MSI naming and explicit rollout defaults

Features

• Custom OpenAI-compatible semantic endpoints now work for activity summarization
• Privacy exclusions are easier to author because plain text patterns match substrings by default
• Raw database snapshots can be exported on a recurring basis to a user-selected folder
• CLI MCP mode lets plugin-based workflows query MemoryLane without the desktop app running
• Windows release assets include both MemoryLane-Setup.exe and MemoryLane-Setup.msi

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Linux and Intel macOS are not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download the latest GitHub release and use either MemoryLane-Setup.exe or MemoryLane-Setup.msi

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.14.2...v0.14.3

MemoryLane v0.14.2

Patch release focused on tray privacy-state reliability and repository maintenance.

What's Changed

• Tray privacy-state reliability - Extracted tray privacy latch logic into a dedicated state module so the "recently paused" state now clears reliably after expiry
• Tray coverage improvements - Added focused tests for tray privacy state transitions and latch expiration behavior
• Repository cleanup and docs - Added a CONTRIBUTING guide, linked it from project docs, and removed obsolete notebook tooling

Features

• Tray privacy status now transitions cleanly from paused to recently paused and back to normal without waiting for extra UI events

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Linux and Intel macOS are not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download MemoryLane-Setup.exe from the latest GitHub release

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.14.1...v0.14.2

MemoryLane v0.14.1

Patch release focused on clearer tray feedback when capture is paused by privacy rules.

What's Changed

• Tray privacy-block status - Added explicit tray tooltip/menu states when capture is paused by privacy rules
• Recent-state latch - Tray now shows a short "recently paused" status after unblocking, making transitions easier to notice
• Release workflow maintenance - Normalized repo-local release skill layout under .agents/skills/release/SKILL.md

Features

• Tray menu now displays Capture paused: privacy rule matched while blocked
• Tray tooltip now reflects active and recently-cleared privacy pauses during capture

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Linux and Intel macOS are not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download MemoryLane-Setup.exe from the latest GitHub release

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.14.0...v0.14.1

MemoryLane v0.14.0

Minor release focused on stronger privacy controls and capture reliability improvements.

What's Changed

• Expanded privacy exclusions - Added app, wildcard URL, and window-title exclusions in Privacy settings to suppress screenshots for sensitive contexts
• Anonymous browsing protection - Capture now pauses for browser private/incognito contexts, with hardened Windows matching to avoid leaks
• Windows app watcher hardening - Refactored watcher internals and improved URL/event handling for better stability and filtering consistency
• Runtime and MCP reliability - Hardened MCP stdio behavior, improved path detection, and added diagnostics/logging for startup and embedding loading

Features

• Privacy settings now support broader exclusion rules for apps, URLs, and window titles
• Browser anonymous-mode and private-window capture suppression is now integrated into the capture pipeline
• Windows watcher pipeline emits stronger app/window metadata for downstream filtering

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Linux and Intel macOS are not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download MemoryLane-Setup.exe from the latest GitHub release

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.13.12...v0.14.0

MemoryLane v0.13.12

Patch release focused on packaging reliability and settings UX cleanup.

What's Changed

• Bundled embedding model - Added model bundling in app builds so semantic capabilities are available without extra manual setup (#71)
• Advanced settings refactor - Split the advanced settings page into focused section components for easier navigation and future maintenance
• Release workflow updates - Refined release/package lock handling and release helper skill instructions

Features

• Local embedding model assets are now shipped with the app package
• Advanced settings UI is organized into clearer sections

Known Issues & Limitations

• Windows OCR still depends on native OCR component availability
• Linux and Intel macOS are not yet officially supported

Installation

• macOS (Apple Silicon): install from the latest GitHub release or via the project install script
• Windows: download MemoryLane-Setup.exe from the latest GitHub release

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.13.11...v0.13.12

MemoryLane v0.13.10

Patch release focused on Slack reply targeting safety.

What's Changed

• Slack owner-message filter - Slack integration now skips owner-authored messages in watched channels, so MemoryLane does not answer the owner's own questions

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.13.9...v0.13.10

MemoryLane v0.13.9

Patch release focused on safer Slack semantic replies and configurable semantic timeout.

What's Changed

• Slack safety guardrails - semantic reply flow now classifies intent before research and blocks password, secret, and sensitive-topic requests
• Configurable LLM timeout - semantic request timeout is now configurable (default 120s) with clearer timeout wording in settings
• Slack setup docs - added Slack app setup guide and a ready-to-paste manifest template
• Maintenance - removed obsolete Claude GitHub Actions workflows

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.13.8...v0.13.9

MemoryLane v0.13.8

Patch release centered on the first beta of Slack integration.

What's Changed

• Slack integration beta - added app settings for Slack polling, approvals, channel selection, and runtime controls
• Semantic Slack replies - MemoryLane can now research recent activity, decide whether context is relevant, and draft short Slack replies before the normal approval flow
• Better Slack guardrails - reply drafting now requires an OpenRouter key and skips low-confidence cases instead of posting weak output
• Minor fixes - cleaned up missing activity video handling and tightened related tests and logging

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.13.7...v0.13.8

MemoryLane v0.13.7

Patch release focused on the packaged MCP server database path.

What's Changed

• Fixed packaged MCP database resolution - the installed Windows app now opens MemoryLane\memorylane.db instead of falling back to the stale dev database path
• Preserved dev MCP behavior - local Electron dev runs still resolve to memorylane\memorylane-dev.db
• Added regression coverage - path resolution now has targeted tests for packaged vs dev Electron executables

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.13.6...v0.13.7

MemoryLane v0.13.6

MemoryLane is a desktop tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server - giving AI assistants like Claude and Cursor memory of what you've been working on. Releases are published from version tags: normal semver tags create standard releases, and suffixed versions such as -beta.1 create prereleases.

What's Changed

• Release assets now use stable filenames - macOS and Windows updater assets now publish under fixed names so latest*.yml and the GitHub release stay aligned
• Release publishing is stricter - the workflow now uploads, verifies, and publishes an explicit asset list for both platforms instead of relying on broad globs
• The activity pipeline naming is simpler - the runtime code was flattened out of src/main/v2 and V2-prefixed types and functions were renamed without changing behavior
• Dead legacy code was removed - unused processor, recorder, and MCP indirection layers were deleted to keep the capture and search stack easier to maintain
• Build metadata was cleaned up - release workflow action updates and package-lock cleanup reduce noise in release builds

Features

• Launch at login
• Persistent capture preference - remembers whether capture should resume after restart or wake
• Unified activity pipeline - runtime path for event/capture ingestion, activity extraction, transformation, and persistence
• Video-first activity understanding - stitched activity clips for richer semantic interpretation with fallback to frame snapshots
• Pattern detection foundation - stores reusable activity patterns for future higher-level context and analysis workflows
• Automatic updates - background update checks with one-click install from the tray menu
• One-command install - curl | sh installer that downloads, installs, and removes quarantine automatically
• Apple notarized - the app is code-signed and Apple-notarized, no Gatekeeper warnings
• Managed API key via Stripe - subscribe and start capturing in seconds, no OpenRouter account needed
• Custom endpoint models - use OpenAI-compatible endpoints, including local runtimes like Ollama
• Multi-screen capture - captures screenshots from all connected displays simultaneously
• Event-driven screen capture - captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• Activity-based processing - groups screenshots into coherent activity sessions for richer summaries
• OCR via macOS Vision - extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization - classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search - vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server - exposes search_context, browse_timeline, and get_event_details tools plus time tracking and recent activity prompts for AI assistants
• One-click integrations - register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture and semantic settings - adjust visual change threshold, typing timeout, scroll timeout, and semantic mode behavior via the UI
• Secure API key storage - uses Electron's safeStorage for encrypted key persistence
• Usage tracking - monitors API requests, token usage, and costs
• Richer activity summaries - improved summary quality for timeline and search context questions
• Windows OCR - native OCR path for the Windows release build
• Windows native capture stack - recorder support for the native watcher and screenshot backends with packaging and integration coverage
• Database export from settings - export local data from the app UI for backup and portability

Known Issues & Limitations

• macOS builds are Apple Silicon only - official macOS release assets currently target ARM64 (.zip and .dmg)
• Linux and Intel macOS not yet officially supported

Installation

curl -fsSL https://raw.githubusercontent.com/deusXmachina-dev/memorylane/main/install.sh | sh

This downloads the latest macOS stable release and installs it to /Applications. No Gatekeeper warnings.

For Windows, download MemoryLane-Setup.exe from the latest GitHub release.

After launching:

1. On macOS, grant Screen Recording permission when prompted
2. On macOS, grant Accessibility permission when prompted
3. Choose how to provide an API key:
   • Subscribe (recommended) - click Subscribe to get a managed key ($10/mo via Stripe)
   • Bring Your Own Key - paste your OpenRouter API key if you already have one
4. Optional: configure a custom model endpoint in settings (for example, a local Ollama endpoint)
5. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.13.5...v0.13.6

MemoryLane v0.13.5

MemoryLane is a desktop tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server - giving AI assistants like Claude and Cursor memory of what you've been working on. Releases are published from version tags: normal semver tags create standard releases, and suffixed versions such as -beta.1 create prereleases.

What's Changed

• Windows is now fully supported - the release pipeline now ships signed Windows installers and updater metadata alongside macOS from the same version tag
• Native Windows screenshot capture is in the main app path - the recorder now uses the dedicated Windows screenshot sidecar with packaging and integration coverage
• Windows OCR and capture flows are more resilient - native OCR handling, recorder integration, and sidecar management were tightened for packaged builds
• Timeline and semantic extraction got minor fixes - activity formatting, transformer behavior, and storage shape were cleaned up for more reliable results
• Release automation is simpler - pushing vX.Y.Z now publishes a standard release, while suffixed versions such as vX.Y.Z-beta.1 publish prereleases automatically

Features

• Launch at login
• Persistent capture preference - remembers whether capture should resume after restart or wake
• V2 activity pipeline - new runtime path for event/capture ingestion, activity extraction, transformation, and persistence
• Video-first activity understanding - stitched activity clips for richer semantic interpretation with fallback to frame snapshots
• Pattern detection foundation - stores reusable activity patterns for future higher-level context and analysis workflows
• Automatic updates - background update checks with one-click install from the tray menu
• One-command install - curl | sh installer that downloads, installs, and removes quarantine automatically
• Apple notarized - the app is code-signed and Apple-notarized, no Gatekeeper warnings
• Managed API key via Stripe - subscribe and start capturing in seconds, no OpenRouter account needed
• Custom endpoint models - use OpenAI-compatible endpoints, including local runtimes like Ollama
• Multi-screen capture - captures screenshots from all connected displays simultaneously
• Event-driven screen capture - captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• Activity-based processing - groups screenshots into coherent activity sessions for richer summaries
• OCR via macOS Vision - extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization - classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search - vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server - exposes search_context, browse_timeline, and get_event_details tools plus time tracking and recent activity prompts for AI assistants
• One-click integrations - register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture and semantic settings - adjust visual change threshold, typing timeout, scroll timeout, and semantic mode behavior via the UI
• Secure API key storage - uses Electron's safeStorage for encrypted key persistence
• Usage tracking - monitors API requests, token usage, and costs
• Richer activity summaries - improved summary quality for timeline and search context questions
• Windows OCR - native OCR path for the Windows release build
• Windows native capture stack - recorder support for the native watcher and screenshot backends with packaging and integration coverage
• Database export from settings - export local data from the app UI for backup and portability

Known Issues & Limitations

• macOS builds are Apple Silicon only - official macOS release assets currently target ARM64 (.zip and .dmg)
• Linux and Intel macOS not yet officially supported

Installation

curl -fsSL https://raw.githubusercontent.com/deusXmachina-dev/memorylane/main/install.sh | sh

This downloads the latest macOS stable release and installs it to /Applications. No Gatekeeper warnings.

For Windows, download the latest MemoryLane Setup *.exe from GitHub Releases.

After launching:

1. On macOS, grant Screen Recording permission when prompted
2. On macOS, grant Accessibility permission when prompted
3. Choose how to provide an API key:
   • Subscribe (recommended) - click Subscribe to get a managed key ($10/mo via Stripe)
   • Bring Your Own Key - paste your OpenRouter API key if you already have one
4. Optional: configure a custom model endpoint in settings (for example, a local Ollama endpoint)
5. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.13.4...v0.13.5

MemoryLane v0.13.4

MemoryLane is a macOS system tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server — giving AI assistants like Claude and Cursor memory of what you've been working on.

What's Changed

• Tray-only auto-start now works end to end (main update) — packaged macOS and Windows builds can register a login item, relaunch hidden in the tray, and stay out of the way at sign-in
• Fresh packaged installs now opt into launch at login automatically — new users start with auto-start enabled by default, and the first packaged run syncs that preference into the OS login item
• Capture state survives relaunches and wake-ups — the app now persists whether capture was enabled, then restores that preference on startup and after power-state resume
• Single-instance startup is more reliable — a second launch now focuses the existing app window instead of creating conflicting tray behavior
• Advanced settings now expose startup controls — added a Launch at login toggle with clearer success/error handling when saving settings

Features

• Launch at login enabled by default on fresh installs — packaged macOS and Windows builds can start automatically and remain hidden in the tray, while still being user-configurable
• Persistent capture preference — remembers whether capture should resume after restart or wake
• V2 activity pipeline — new runtime path for event/capture ingestion, activity extraction, transformation, and persistence
• Video-first activity understanding — stitched activity clips for richer semantic interpretation with fallback to frame snapshots
• Pattern detection foundation — stores reusable activity patterns for future higher-level context and analysis workflows
• Automatic updates — background update checks with one-click install from the tray menu
• One-command install — curl | sh installer that downloads, installs, and removes quarantine automatically
• Apple notarized — the app is code-signed and Apple-notarized, no Gatekeeper warnings
• Managed API key via Stripe — subscribe and start capturing in seconds, no OpenRouter account needed
• Custom endpoint models — use OpenAI-compatible endpoints, including local runtimes like Ollama
• Multi-screen capture — captures screenshots from all connected displays simultaneously
• Event-driven screen capture — captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• Activity-based processing — groups screenshots into coherent activity sessions for richer summaries
• OCR via macOS Vision — extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization — classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search — vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server — exposes search_context, browse_timeline, and get_event_details tools plus time tracking and recent activity prompts for AI assistants
• One-click integrations — register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture and semantic settings — adjust visual change threshold, typing timeout, scroll timeout, and semantic mode behavior via the UI
• Secure API key storage — uses Electron's safeStorage for encrypted key persistence
• Usage tracking — monitors API requests, token usage, and costs
• Richer activity summaries — improved summary quality for timeline and search context questions
• Windows OCR (preview) — native OCR path available for Windows preview setups
• Windows app watcher integration (preview) — recorder support for the native watcher backend with unit/e2e coverage and build packaging
• Database export from settings — export local data from the app UI for backup and portability

Known Issues & Limitations

• macOS ARM64 release artifact only — official release assets are currently Apple Silicon macOS (.zip and .dmg)
• Windows support is preview quality — native OCR is available, but some OS-specific UX and setup polish are still in progress
• Linux and Intel macOS not yet officially supported

Installation

curl -fsSL https://raw.githubusercontent.com/deusXmachina-dev/memorylane/main/install.sh | sh

This downloads the latest release and installs it to /Applications. No Gatekeeper warnings.

After launching:

1. Grant Screen Recording permission when prompted
2. Grant Accessibility permission when prompted
3. Choose how to provide an API key:
   • Subscribe (recommended) — click Subscribe to get a managed key ($10/mo via Stripe)
   • Bring Your Own Key — paste your OpenRouter API key if you already have one
4. Optional: configure a custom model endpoint in settings (for example, a local Ollama endpoint)
5. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.13.2...v0.13.4

MemoryLane v0.13.2

MemoryLane is a macOS system tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server — giving AI assistants like Claude and Cursor memory of what you've been working on.

What's Changed

• v2 snapshots now behave closer to v1 (main update) — reused v1 snapshot settings and tightened frame retention rules so fallback snapshots better match the proven v1 capture behavior
• More accurate activity boundaries and frame timing — fixed event/window boundary start timestamps and callback-time frame timestamps to improve temporal consistency in produced activities
• Sharper snapshot boundary selection — constrained boundary anchor selection by direction and added tests around late events and boundary edges
• Settings and app polish — added semantic mode selection in advanced settings and fixed macOS tray auto-update quit behavior

Features

• V2 activity pipeline — new runtime path for event/capture ingestion, activity extraction, transformation, and persistence
• Video-first activity understanding — stitched activity clips for richer semantic interpretation with fallback to frame snapshots
• Pattern detection foundation — stores reusable activity patterns for future higher-level context and analysis workflows
• Automatic updates — background update checks with one-click install from the tray menu
• One-command install — curl | sh installer that downloads, installs, and removes quarantine automatically
• Apple notarized — the app is code-signed and Apple-notarized, no Gatekeeper warnings
• Managed API key via Stripe — subscribe and start capturing in seconds, no OpenRouter account needed
• Custom endpoint models — use OpenAI-compatible endpoints, including local runtimes like Ollama
• Multi-screen capture — captures screenshots from all connected displays simultaneously
• Event-driven screen capture — captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• Activity-based processing — groups screenshots into coherent activity sessions for richer summaries
• OCR via macOS Vision — extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization — classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search — vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server — exposes search_context, browse_timeline, and get_event_details tools plus time tracking and recent activity prompts for AI assistants
• One-click integrations — register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture and semantic settings — adjust visual change threshold, typing timeout, scroll timeout, and semantic mode behavior via the UI
• Secure API key storage — uses Electron's safeStorage for encrypted key persistence
• Usage tracking — monitors API requests, token usage, and costs
• Richer activity summaries — improved summary quality for timeline and search context questions
• Windows OCR (preview) — native OCR path available for Windows preview setups
• Windows app watcher integration (preview) — recorder support for the native watcher backend with unit/e2e coverage and build packaging
• Database export from settings — export local data from the app UI for backup and portability

Known Issues & Limitations

• macOS ARM64 release artifact only — official release assets are currently Apple Silicon macOS (.zip and .dmg)
• Windows support is preview quality — native OCR is available, but some OS-specific UX and setup polish are still in progress
• Linux and Intel macOS not yet officially supported

Installation

curl -fsSL https://raw.githubusercontent.com/deusXmachina-dev/memorylane/main/install.sh | sh

This downloads the latest release and installs it to /Applications. No Gatekeeper warnings.

After launching:

1. Grant Screen Recording permission when prompted
2. Grant Accessibility permission when prompted
3. Choose how to provide an API key:
   • Subscribe (recommended) — click Subscribe to get a managed key ($10/mo via Stripe)
   • Bring Your Own Key — paste your OpenRouter API key if you already have one
4. Optional: configure a custom model endpoint in settings (for example, a local Ollama endpoint)
5. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.13.1...v0.13.2

MemoryLane v0.13.1

MemoryLane is a macOS system tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server — giving AI assistants like Claude and Cursor memory of what you've been working on.

What's Changed

• Better semantic prompt quality (main update) — aligned the v2 semantic prompt with the proven v1 phrasing to improve activity summary recall and consistency
• Faster snapshot fallback cadence — reduced the default v2 semantic snapshot spacing to 5 seconds for denser visual context when video mode falls back
• Accurate usage cost tracking for video defaults — added missing pricing entries for default v2 semantic video models so reported token costs are correct

Features

• V2 activity pipeline — new runtime path for event/capture ingestion, activity extraction, transformation, and persistence
• Video-first activity understanding — stitched activity clips for richer semantic interpretation with fallback to frame snapshots
• Pattern detection foundation — stores reusable activity patterns for future higher-level context and analysis workflows
• Automatic updates — background update checks with one-click install from the tray menu
• One-command install — curl | sh installer that downloads, installs, and removes quarantine automatically
• Apple notarized — the app is code-signed and Apple-notarized, no Gatekeeper warnings
• Managed API key via Stripe — subscribe and start capturing in seconds, no OpenRouter account needed
• Custom endpoint models — use OpenAI-compatible endpoints, including local runtimes like Ollama
• Multi-screen capture — captures screenshots from all connected displays simultaneously
• Event-driven screen capture — captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• Activity-based processing — groups screenshots into coherent activity sessions for richer summaries
• OCR via macOS Vision — extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization — classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search — vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server — exposes search_context, browse_timeline, and get_event_details tools plus time tracking and recent activity prompts for AI assistants
• One-click integrations — register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture settings — adjust visual change threshold, typing timeout, scroll timeout via the UI
• Secure API key storage — uses Electron's safeStorage for encrypted key persistence
• Usage tracking — monitors API requests, token usage, and costs
• Richer activity summaries — improved summary quality for timeline and search context questions
• Windows OCR (preview) — native OCR path available for Windows preview setups
• Windows app watcher integration (preview) — recorder support for the native watcher backend with unit/e2e coverage and build packaging
• Database export from settings — export local data from the app UI for backup and portability

Known Issues & Limitations

• macOS ARM64 release artifact only — official release assets are currently Apple Silicon macOS (.zip and .dmg)
• Windows support is preview quality — native OCR is available, but some OS-specific UX and setup polish are still in progress
• Linux and Intel macOS not yet officially supported

Installation

curl -fsSL https://raw.githubusercontent.com/deusXmachina-dev/memorylane/main/install.sh | sh

This downloads the latest release and installs it to /Applications. No Gatekeeper warnings.

After launching:

1. Grant Screen Recording permission when prompted
2. Grant Accessibility permission when prompted
3. Choose how to provide an API key:
   • Subscribe (recommended) — click Subscribe to get a managed key ($10/mo via Stripe)
   • Bring Your Own Key — paste your OpenRouter API key if you already have one
4. Optional: configure a custom model endpoint in settings (for example, a local Ollama endpoint)
5. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.13.0...v0.13.1

MemoryLane v0.13.0

MemoryLane is a macOS system tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server — giving AI assistants like Claude and Cursor memory of what you've been working on.

What's Changed

• Windows app watcher backend (preview) — added a native Rust sidecar for window/activity monitoring with recorder display routing and packaging integration
• Database export and backup — added a UI flow to export local database contents as a zip backup (closes #48)
• Capture and stitching reliability improvements — removed hardcoded FFmpeg thread limits, improved screenshot capture internals, and expanded recorder/video test coverage
• Tray UX improvement — added a full-size inverted tray icon asset for better menu bar visibility in more themes

Features

• V2 activity pipeline — new runtime path for event/capture ingestion, activity extraction, transformation, and persistence
• Video-first activity understanding — stitched activity clips for richer semantic interpretation with fallback to frame snapshots
• Pattern detection foundation — stores reusable activity patterns for future higher-level context and analysis workflows
• Automatic updates — background update checks with one-click install from the tray menu
• One-command install — curl | sh installer that downloads, installs, and removes quarantine automatically
• Apple notarized — the app is code-signed and Apple-notarized, no Gatekeeper warnings
• Managed API key via Stripe — subscribe and start capturing in seconds, no OpenRouter account needed
• Custom endpoint models — use OpenAI-compatible endpoints, including local runtimes like Ollama
• Multi-screen capture — captures screenshots from all connected displays simultaneously
• Event-driven screen capture — captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• Activity-based processing — groups screenshots into coherent activity sessions for richer summaries
• OCR via macOS Vision — extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization — classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search — vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server — exposes search_context, browse_timeline, and get_event_details tools plus time tracking and recent activity prompts for AI assistants
• One-click integrations — register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture settings — adjust visual change threshold, typing timeout, scroll timeout via the UI
• Secure API key storage — uses Electron's safeStorage for encrypted key persistence
• Usage tracking — monitors API requests, token usage, and costs
• Richer activity summaries — improved summary quality for timeline and search context questions
• Windows OCR (preview) — native OCR path available for Windows preview setups
• Windows app watcher integration (preview) — recorder support for the native watcher backend with unit/e2e coverage and build packaging
• Database export from settings — export local data from the app UI for backup and portability

Known Issues & Limitations

• macOS ARM64 release artifact only — official release assets are currently Apple Silicon macOS (.zip and .dmg)
• Windows support is preview quality — native OCR is available, but some OS-specific UX and setup polish are still in progress
• Linux and Intel macOS not yet officially supported

Installation

curl -fsSL https://raw.githubusercontent.com/deusXmachina-dev/memorylane/main/install.sh | sh

This downloads the latest release and installs it to /Applications. No Gatekeeper warnings.

After launching:

1. Grant Screen Recording permission when prompted
2. Grant Accessibility permission when prompted
3. Choose how to provide an API key:
   • Subscribe (recommended) — click Subscribe to get a managed key ($10/mo via Stripe)
   • Bring Your Own Key — paste your OpenRouter API key if you already have one
4. Optional: configure a custom model endpoint in settings (for example, a local Ollama endpoint)
5. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.12.1...v0.13.0

MemoryLane v0.12.1

MemoryLane is a macOS system tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server — giving AI assistants like Claude and Cursor memory of what you've been working on.

What's Changed

• Fixed auto updater — resolved an issue where the automatic update mechanism could fail to download and apply updates correctly

Features

• V2 activity pipeline — new runtime path for event/capture ingestion, activity extraction, transformation, and persistence
• Video-first activity understanding — stitched activity clips for richer semantic interpretation with fallback to frame snapshots
• Pattern detection foundation — stores reusable activity patterns for future higher-level context and analysis workflows
• Automatic updates — background update checks with one-click install from the tray menu
• One-command install — curl | sh installer that downloads, installs, and removes quarantine automatically
• Apple notarized — the app is code-signed and Apple-notarized, no Gatekeeper warnings
• Managed API key via Stripe — subscribe and start capturing in seconds, no OpenRouter account needed
• Custom endpoint models — use OpenAI-compatible endpoints, including local runtimes like Ollama
• Multi-screen capture — captures screenshots from all connected displays simultaneously
• Event-driven screen capture — captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• Activity-based processing — groups screenshots into coherent activity sessions for richer summaries
• OCR via macOS Vision — extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization — classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search — vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server — exposes search_context, browse_timeline, and get_event_details tools plus time tracking and recent activity prompts for AI assistants
• One-click integrations — register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture settings — adjust visual change threshold, typing timeout, scroll timeout via the UI
• Secure API key storage — uses Electron's safeStorage for encrypted key persistence
• Usage tracking — monitors API requests, token usage, and costs
• Richer activity summaries — improved summary quality for timeline and search context questions
• Windows OCR (preview) — native OCR path available for Windows preview setups

Known Issues & Limitations

• macOS ARM64 release artifact only — official release assets are currently Apple Silicon macOS (.zip and .dmg)
• Windows support is preview quality — native OCR is available, but some OS-specific UX and setup polish are still in progress
• Linux and Intel macOS not yet officially supported

Installation

curl -fsSL https://raw.githubusercontent.com/deusXmachina-dev/memorylane/main/install.sh | sh

This downloads the latest release and installs it to /Applications. No Gatekeeper warnings.

After launching:

1. Grant Screen Recording permission when prompted
2. Grant Accessibility permission when prompted
3. Choose how to provide an API key:
   • Subscribe (recommended) — click Subscribe to get a managed key ($10/mo via Stripe)
   • Bring Your Own Key — paste your OpenRouter API key if you already have one
4. Optional: configure a custom model endpoint in settings (for example, a local Ollama endpoint)
5. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.12.0...v0.12.1

MemoryLane v0.12.0

MemoryLane is a macOS system tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server — giving AI assistants like Claude and Cursor memory of what you've been working on.

What's Changed

• V2 runtime pipeline — introduced a full v2 activity runtime with durable streams, coordinated extraction/production stages, and richer activity lifecycle handling
• Video-first semantic processing — added ffmpeg-based activity video stitching and a semantic service that prefers stitched media with robust snapshot fallback behavior
• Pattern intelligence and MCP improvements — added pattern detection infrastructure, storage/repository support, and MCP tool/prompt enhancements for better context retrieval
• Native capture hardening on macOS — improved v2 screenshot capture via ScreenCaptureKit/CoreGraphics paths, focused-window display tracking, and transient-failure tolerance
• Reliability, performance, and test depth — reduced capture/media footprint, improved cleanup/shutdown semantics, and added broad unit/integration coverage across the v2 stack

Features

• V2 activity pipeline — new runtime path for event/capture ingestion, activity extraction, transformation, and persistence
• Video-first activity understanding — stitched activity clips for richer semantic interpretation with fallback to frame snapshots
• Pattern detection foundation — stores reusable activity patterns for future higher-level context and analysis workflows
• Automatic updates — background update checks with one-click install from the tray menu
• One-command install — curl | sh installer that downloads, installs, and removes quarantine automatically
• Apple notarized — the app is code-signed and Apple-notarized, no Gatekeeper warnings
• Managed API key via Stripe — subscribe and start capturing in seconds, no OpenRouter account needed
• Custom endpoint models — use OpenAI-compatible endpoints, including local runtimes like Ollama
• Multi-screen capture — captures screenshots from all connected displays simultaneously
• Event-driven screen capture — captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• Activity-based processing — groups screenshots into coherent activity sessions for richer summaries
• OCR via macOS Vision — extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization — classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search — vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server — exposes search_context, browse_timeline, and get_event_details tools plus time tracking and recent activity prompts for AI assistants
• One-click integrations — register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture settings — adjust visual change threshold, typing timeout, scroll timeout via the UI
• Secure API key storage — uses Electron's safeStorage for encrypted key persistence
• Usage tracking — monitors API requests, token usage, and costs
• Richer activity summaries — improved summary quality for timeline and search context questions
• Windows OCR (preview) — native OCR path available for Windows preview setups

Known Issues & Limitations

• macOS ARM64 release artifact only — official release assets are currently Apple Silicon macOS (.zip and .dmg)
• Windows support is preview quality — native OCR is available, but some OS-specific UX and setup polish are still in progress
• Linux and Intel macOS not yet officially supported

Installation

curl -fsSL https://raw.githubusercontent.com/deusXmachina-dev/memorylane/main/install.sh | sh

This downloads the latest release and installs it to /Applications. No Gatekeeper warnings.

After launching:

1. Grant Screen Recording permission when prompted
2. Grant Accessibility permission when prompted
3. Choose how to provide an API key:
   • Subscribe (recommended) — click Subscribe to get a managed key ($10/mo via Stripe)
   • Bring Your Own Key — paste your OpenRouter API key if you already have one
4. Optional: configure a custom model endpoint in settings (for example, a local Ollama endpoint)
5. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.11.0...v0.12.0

MemoryLane v0.11.0

MemoryLane is a macOS system tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server — giving AI assistants like Claude and Cursor memory of what you've been working on.

What's Changed

• Native Swift screenshot module — added a new native macOS screenshot capture path using Swift, replacing the Electron desktopCapturer dependency for faster, lower-overhead screen capture
• Fixed-interval ScreenCapturer — new ScreenCapturer class that supports fixed-interval frame capture alongside the existing event-driven mode
• Configurable capture settings UI — adjust visual change threshold, typing timeout, and scroll timeout directly from the main window (#28, closes #33)
• URL parsing fix — fixed URL parsing in the bundled app that could cause failures in production builds
• Internal improvements — better bundleId separation for dev/production builds, updated MCP prompt quality, increased max history size, added migration tooling and architecture documentation

Features

• Automatic updates — background update checks with one-click install from the tray menu
• One-command install — curl | sh installer that downloads, installs, and removes quarantine automatically
• Apple notarized — the app is code-signed and Apple-notarized, no Gatekeeper warnings
• Managed API key via Stripe — subscribe and start capturing in seconds, no OpenRouter account needed
• Custom endpoint models — use OpenAI-compatible endpoints, including local runtimes like Ollama
• Multi-screen capture — captures screenshots from all connected displays simultaneously
• Event-driven screen capture — captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• Activity-based processing — groups screenshots into coherent activity sessions for richer summaries
• OCR via macOS Vision — extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization — classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search — vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server — exposes search_context, browse_timeline, and get_event_details tools plus time tracking and recent activity prompts for AI assistants
• One-click integrations — register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture settings — adjust visual change threshold, typing timeout, scroll timeout via the UI
• Secure API key storage — uses Electron's safeStorage for encrypted key persistence
• Usage tracking — monitors API requests, token usage, and costs
• Richer activity summaries — improved summary quality for timeline and search context questions
• Windows OCR (preview) — native OCR path available for Windows preview setups

Known Issues & Limitations

• macOS ARM64 release artifact only — official release assets are currently Apple Silicon macOS (.zip and .dmg)
• Windows support is preview quality — native OCR is available, but some OS-specific UX and setup polish are still in progress
• Linux and Intel macOS not yet officially supported

Installation

curl -fsSL https://raw.githubusercontent.com/deusXmachina-dev/memorylane/main/install.sh | sh

This downloads the latest release and installs it to /Applications. No Gatekeeper warnings.

After launching:

1. Grant Screen Recording permission when prompted
2. Grant Accessibility permission when prompted
3. Choose how to provide an API key:
   • Subscribe (recommended) — click Subscribe to get a managed key ($10/mo via Stripe)
   • Bring Your Own Key — paste your OpenRouter API key if you already have one
4. Optional: configure a custom model endpoint in settings (for example, a local Ollama endpoint)
5. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.10.0...v0.11.0

MemoryLane v0.10.0

MemoryLane is a macOS system tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server — giving AI assistants like Claude and Cursor memory of what you've been working on.

What's Changed

• Automatic updates — the app now checks for new releases on launch and every 4 hours, downloads updates in the background, and shows an "Install Update Now" option in the tray menu when ready
• GitHub publish config — electron-builder.yml now includes a GitHub publish provider, enabling electron-updater to detect and download new versions from GitHub Releases
• Simplified tray menu — removed the per-session API usage stats submenu for a cleaner tray experience; database stats remain available

Features

• Automatic updates — background update checks with one-click install from the tray menu
• One-command install — curl | sh installer that downloads, installs, and removes quarantine automatically
• Apple notarized — the app is code-signed and Apple-notarized, no Gatekeeper warnings
• Managed API key via Stripe — subscribe and start capturing in seconds, no OpenRouter account needed
• Custom endpoint models — use OpenAI-compatible endpoints, including local runtimes like Ollama
• Multi-screen capture — captures screenshots from all connected displays simultaneously
• Event-driven screen capture — captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• Activity-based processing — groups screenshots into coherent activity sessions for richer summaries
• OCR via macOS Vision — extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization — classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search — vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server — exposes search_context, browse_timeline, and get_event_details tools plus time tracking and recent activity prompts for AI assistants
• One-click integrations — register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture settings — adjust visual change threshold, typing timeout, scroll timeout via the UI
• Secure API key storage — uses Electron's safeStorage for encrypted key persistence
• Usage tracking — monitors API requests, token usage, and costs
• Richer activity summaries — improved summary quality for timeline and search context questions
• Windows OCR (preview) — native OCR path available for Windows preview setups

Known Issues & Limitations

• macOS ARM64 release artifact only — official release assets are currently Apple Silicon macOS (.zip and .dmg)
• Windows support is preview quality — native OCR is available, but some OS-specific UX and setup polish are still in progress
• Linux and Intel macOS not yet officially supported

Installation

curl -fsSL https://raw.githubusercontent.com/deusXmachina-dev/memorylane/main/install.sh | sh

This downloads the latest release and installs it to /Applications. No Gatekeeper warnings.

After launching:

1. Grant Screen Recording permission when prompted
2. Grant Accessibility permission when prompted
3. Choose how to provide an API key:
   • Subscribe (recommended) — click Subscribe to get a managed key ($10/mo via Stripe)
   • Bring Your Own Key — paste your OpenRouter API key if you already have one
4. Optional: configure a custom model endpoint in settings (for example, a local Ollama endpoint)
5. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.9.1...v0.10.0

MemoryLane v0.9.1

MemoryLane is a macOS system tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server — giving AI assistants like Claude and Cursor memory of what you've been working on.

What's Changed

• Fixed model cache path for packaged builds — the Transformers.js model cache now uses an absolute path under the app's data directory, fixing an ENOENT crash when macOS launches the packaged app with cwd=/
• More accurate OCR — switched the macOS Vision recognition mode from fast to accurate for higher-quality text extraction

Features

• One-command install — curl | sh installer that downloads, installs, and removes quarantine automatically
• Apple notarized — the app is code-signed and Apple-notarized, no Gatekeeper warnings
• Managed API key via Stripe — subscribe and start capturing in seconds, no OpenRouter account needed
• Custom endpoint models — use OpenAI-compatible endpoints, including local runtimes like Ollama
• Multi-screen capture — captures screenshots from all connected displays simultaneously
• Event-driven screen capture — captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• Activity-based processing — groups screenshots into coherent activity sessions for richer summaries
• OCR via macOS Vision — extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization — classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search — vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server — exposes search_context, browse_timeline, and get_event_details tools plus time tracking and recent activity prompts for AI assistants
• One-click integrations — register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture settings — adjust visual change threshold, typing timeout, scroll timeout via the UI
• Secure API key storage — uses Electron's safeStorage for encrypted key persistence
• Usage tracking — monitors API requests, token usage, and costs
• Richer activity summaries — improved summary quality for timeline and search context questions
• Windows OCR (preview) — native OCR path available for Windows preview setups

Known Issues & Limitations

• macOS ARM64 release artifact only — official release assets are currently Apple Silicon macOS (.zip and .dmg)
• Windows support is preview quality — native OCR is available, but some OS-specific UX and setup polish are still in progress
• Linux and Intel macOS not yet officially supported

Installation

curl -fsSL https://raw.githubusercontent.com/deusXmachina-dev/memorylane/main/install.sh | sh

This downloads the latest release and installs it to /Applications. No Gatekeeper warnings.

After launching:

1. Grant Screen Recording permission when prompted
2. Grant Accessibility permission when prompted
3. Choose how to provide an API key:
   • Subscribe (recommended) — click Subscribe to get a managed key ($10/mo via Stripe)
   • Bring Your Own Key — paste your OpenRouter API key if you already have one
4. Optional: configure a custom model endpoint in settings (for example, a local Ollama endpoint)
5. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.9.0...v0.9.1

MemoryLane v0.9.0

MemoryLane is a macOS system tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server — giving AI assistants like Claude and Cursor memory of what you've been working on.

What's Changed

• Activity-based processing — screenshots are now grouped into coherent activity sessions via a new ActivityManager, replacing individual screenshot processing with richer, context-aware summaries
• Pure event-driven capture — removed periodic screenshots entirely; capture is now triggered exclusively by user interactions (clicks, typing, scrolling, app switches) and visual changes
• Click capture trigger — mouse clicks now trigger screenshots, improving activity transition detection
• Improved OCR pipeline — polished OCR firing sequences and Swift sidecar for more accurate text extraction
• Better summarization prompts — refined prompts and removed underperforming models for higher-quality activity summaries
• Processing queue — new serial processing queue ensures ordered, reliable screenshot processing
• MCP response improvements — enhanced formatting and richer context in search_context and browse_timeline tool responses
• Data migration — existing screenshot data is automatically migrated into the new activity-based schema on upgrade
• Bug fixes — resolved race conditions in processing, cold start issues, and Transformers.js compatibility problems

Features

• One-command install — curl | sh installer that downloads, installs, and removes quarantine automatically
• Apple notarized — the app is code-signed and Apple-notarized, no Gatekeeper warnings
• Managed API key via Stripe — subscribe and start capturing in seconds, no OpenRouter account needed
• Custom endpoint models — use OpenAI-compatible endpoints, including local runtimes like Ollama
• Multi-screen capture — captures screenshots from all connected displays simultaneously
• Event-driven screen capture — captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• Activity-based processing — groups screenshots into coherent activity sessions for richer summaries
• OCR via macOS Vision — extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization — classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search — vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server — exposes search_context, browse_timeline, and get_event_details tools plus time tracking and recent activity prompts for AI assistants
• One-click integrations — register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture settings — adjust visual change threshold, typing timeout, scroll timeout via the UI
• Secure API key storage — uses Electron's safeStorage for encrypted key persistence
• Usage tracking — monitors API requests, token usage, and costs
• Richer activity summaries — improved summary quality for timeline and search context questions
• Windows OCR (preview) — native OCR path available for Windows preview setups

Known Issues & Limitations

• macOS ARM64 release artifact only — official release assets are currently Apple Silicon macOS (.zip and .dmg)
• Windows support is preview quality — native OCR is available, but some OS-specific UX and setup polish are still in progress
• Linux and Intel macOS not yet officially supported

Installation

curl -fsSL https://raw.githubusercontent.com/deusXmachina-dev/memorylane/main/install.sh | sh

This downloads the latest release and installs it to /Applications. No Gatekeeper warnings.

After launching:

1. Grant Screen Recording permission when prompted
2. Grant Accessibility permission when prompted
3. Choose how to provide an API key:
   • Subscribe (recommended) — click Subscribe to get a managed key ($10/mo via Stripe)
   • Bring Your Own Key — paste your OpenRouter API key if you already have one
4. Optional: configure a custom model endpoint in settings (for example, a local Ollama endpoint)
5. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.8.0...v0.9.0

MemoryLane v0.8.0

MemoryLane is a macOS system tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server — giving AI assistants like Claude and Cursor memory of what you've been working on.

What's Changed

• Custom endpoint model support (local models included) — configure an OpenAI-compatible endpoint in the app and run summarization through your own hosted or local models (for example Ollama-served open weights)
• End-to-end custom endpoint integration — new shared types, persistent endpoint settings, preload/main-process IPC wiring, and runtime model client selection in semantic classification
• Custom endpoint UI in the main window — new settings section for endpoint/model configuration, including a collapsible layout for easier day-to-day use
• Improved app context detection — recorder now monitors title changes to improve activity transitions and timeline quality

Features

• One-command install — curl | sh installer that downloads, installs, and removes quarantine automatically
• Apple notarized — the app is code-signed and Apple-notarized, no Gatekeeper warnings
• Managed API key via Stripe — subscribe and start capturing in seconds, no OpenRouter account needed
• Custom endpoint models — use OpenAI-compatible endpoints, including local runtimes like Ollama
• Multi-screen capture — captures screenshots from all connected displays simultaneously
• Event-driven screen capture — captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• OCR via macOS Vision — extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization — classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search — vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server — exposes search_context, browse_timeline, and get_event_details tools plus time tracking and recent activity prompts for AI assistants
• One-click integrations — register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture settings — adjust visual change threshold, typing timeout, scroll timeout via the UI
• Secure API key storage — uses Electron's safeStorage for encrypted key persistence
• Usage tracking — monitors API requests, token usage, and costs
• Richer activity summaries — improved summary quality for timeline and search context questions
• Windows OCR (preview) — native OCR path available for Windows preview setups

Known Issues & Limitations

• macOS ARM64 release artifact only — official release assets are currently Apple Silicon macOS (.zip and .dmg)
• Windows support is preview quality — native OCR is available, but some OS-specific UX and setup polish are still in progress
• Linux and Intel macOS not yet officially supported

Installation

curl -fsSL https://raw.githubusercontent.com/deusXmachina-dev/memorylane/main/install.sh | sh

This downloads the latest release and installs it to /Applications. No Gatekeeper warnings.

After launching:

1. Grant Screen Recording permission when prompted
2. Grant Accessibility permission when prompted
3. Choose how to provide an API key:
   • Subscribe (recommended) — click Subscribe to get a managed key ($10/mo via Stripe)
   • Bring Your Own Key — paste your OpenRouter API key if you already have one
4. Optional: configure a custom model endpoint in settings (for example, a local Ollama endpoint)
5. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.7.0...v0.8.0

MemoryLane v0.7.0

MemoryLane is a macOS system tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server — giving AI assistants like Claude and Cursor memory of what you've been working on.

What's Changed

• Richer activity summaries — semantic classification now produces more detailed, useful summaries of what you were doing on screen
• Summary-first MCP guidance — tool prompts now prioritize activity summaries for "what was I doing?" questions and reserve OCR for exact recall
• Event-based app change monitoring — capture timing now reacts more precisely to active app transitions
• Power usage improvements — background monitoring was tuned to reduce unnecessary wakeups and improve efficiency
• Windows preview improvements — native Windows OCR support and additional platform groundwork landed in this release

Features

• One-command install — curl | sh installer that downloads, installs, and removes quarantine automatically
• Apple notarized — the app is code-signed and Apple-notarized, no Gatekeeper warnings
• Managed API key via Stripe — subscribe and start capturing in seconds, no OpenRouter account needed
• Multi-screen capture — captures screenshots from all connected displays simultaneously
• Event-driven screen capture — captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• OCR via macOS Vision — extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization — classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search — vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server — exposes search_context, browse_timeline, and get_event_details tools plus time tracking and recent activity prompts for AI assistants
• One-click integrations — register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture settings — adjust visual change threshold, typing timeout, scroll timeout via the UI
• Secure API key storage — uses Electron's safeStorage for encrypted key persistence
• Usage tracking — monitors API requests, token usage, and costs
• Richer activity summaries — improved summary quality for timeline and search context questions
• Windows OCR (preview) — native OCR path available for Windows preview setups

Known Issues & Limitations

• macOS ARM64 release artifact only — official release assets are currently Apple Silicon macOS (.zip and .dmg)
• Windows support is preview quality — native OCR is available, but some OS-specific UX and setup polish are still in progress
• Linux and Intel macOS not yet officially supported

Installation

curl -fsSL https://raw.githubusercontent.com/deusXmachina-dev/memorylane/main/install.sh | sh

This downloads the latest release and installs it to /Applications. No Gatekeeper warnings.

After launching:

1. Grant Screen Recording permission when prompted
2. Grant Accessibility permission when prompted
3. Choose how to provide an API key:
   • Subscribe (recommended) — click Subscribe to get a managed key ($10/mo via Stripe)
   • Bring Your Own Key — paste your OpenRouter API key if you already have one
4. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

https://github.com/deusXmachina-dev/memorylane/compare/v0.6.2...v0.7.0

MemoryLane v0.6.2

MemoryLane is a macOS system tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server — giving AI assistants like Claude and Cursor memory of what you've been working on.

What's Changed

• Hidden dock icon — the app no longer appears in the macOS dock, providing a cleaner tray-only experience

Features

• One-command install — curl | sh installer that downloads, installs, and removes quarantine automatically
• Apple notarized — the app is code-signed and Apple-notarized, no Gatekeeper warnings
• Managed API key via Stripe — subscribe and start capturing in seconds, no OpenRouter account needed
• Multi-screen capture — captures screenshots from all connected displays simultaneously
• Event-driven screen capture — captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• OCR via macOS Vision — extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization — classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search — vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server — exposes search_context, browse_timeline, and get_event_details tools plus time tracking and recent activity prompts for AI assistants
• One-click integrations — register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture settings — adjust visual change threshold, typing timeout, scroll timeout via the UI
• Secure API key storage — uses Electron's safeStorage for encrypted key persistence
• Usage tracking — monitors API requests, token usage, and costs

Known Issues & Limitations

• macOS ARM64 only — this release is Apple Silicon only; Intel Mac, Windows, and Linux builds are not yet available

Installation

curl -fsSL https://raw.githubusercontent.com/deusXmachina-dev/memorylane/main/install.sh | sh

This downloads the latest release and installs it to /Applications. No Gatekeeper warnings.

After launching:

1. Grant Screen Recording permission when prompted
2. Grant Accessibility permission when prompted
3. Choose how to provide an API key:
   • Subscribe (recommended) — click Subscribe to get a managed key ($20/mo via Stripe)
   • Bring Your Own Key — paste your OpenRouter API key if you already have one
4. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

https://github.com/deusXmachina-dev/memorylane/commits/v0.6.2

MemoryLane v0.6.1

MemoryLane is a macOS system tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server — giving AI assistants like Claude and Cursor memory of what you've been working on.

What's Changed

• Pre-compiled Swift OCR binary — the OCR sidecar is now compiled during the build step and shipped as a native binary, instead of being interpreted via swift at runtime. This fixes OCR failures on machines without the Swift toolchain and improves OCR startup performance.

Features

• One-command install — curl | sh installer that downloads, installs, and removes quarantine automatically
• Apple notarized — the app is code-signed and Apple-notarized, no Gatekeeper warnings
• Managed API key via Stripe — subscribe and start capturing in seconds, no OpenRouter account needed
• Multi-screen capture — captures screenshots from all connected displays simultaneously
• Event-driven screen capture — captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• OCR via macOS Vision — extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization — classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search — vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server — exposes search_context, browse_timeline, and get_event_details tools plus time tracking and recent activity prompts for AI assistants
• One-click integrations — register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture settings — adjust visual change threshold, typing timeout, scroll timeout via the UI
• Secure API key storage — uses Electron's safeStorage for encrypted key persistence
• Usage tracking — monitors API requests, token usage, and costs

Known Issues & Limitations

• macOS ARM64 only — this release is Apple Silicon only; Intel Mac, Windows, and Linux builds are not yet available

Installation

curl -fsSL https://raw.githubusercontent.com/deusXmachina-dev/memorylane/main/install.sh | sh

This downloads the latest release and installs it to /Applications. No Gatekeeper warnings.

After launching:

1. Grant Screen Recording permission when prompted
2. Grant Accessibility permission when prompted
3. Choose how to provide an API key:
   • Subscribe (recommended) — click Subscribe to get a managed key ($20/mo via Stripe)
   • Bring Your Own Key — paste your OpenRouter API key if you already have one
4. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

https://github.com/deusXmachina-dev/memorylane/commits/v0.6.1

MemoryLane v0.6.0

MemoryLane is a macOS system tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server — giving AI assistants like Claude and Cursor memory of what you've been working on.

What's Changed

• MCP server no longer blocks app startup — the MCP server now runs as a separate Node process instead of inside the Electron main process, so the app launches instantly regardless of how many MCP clients are connected
• Snapshot timing fix — pending events are now captured at screenshot time rather than processing time, improving the accuracy of activity timestamps
• Debug mode for LLM pipeline — added a debug mode that lets you inspect how each screenshot flows through the LLM summarization pipeline
• Database exploration notebook — added a Jupyter notebook (tools/explore-db.ipynb) for ad-hoc exploration of the local SQLite database
• Stable release artifact names — release assets now use versionless filenames so download URLs always resolve to the latest version
• Claude Code CI workflow — added a GitHub Actions workflow for Claude Code review

Features

• One-command install — curl | sh installer that downloads, installs, and removes quarantine automatically
• Apple notarized — the app is code-signed and Apple-notarized, no Gatekeeper warnings
• Managed API key via Stripe — subscribe and start capturing in seconds, no OpenRouter account needed
• Multi-screen capture — captures screenshots from all connected displays simultaneously
• Event-driven screen capture — captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• OCR via macOS Vision — extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization — classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search — vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server — exposes search_context, browse_timeline, and get_event_details tools plus time tracking and recent activity prompts for AI assistants
• One-click integrations — register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture settings — adjust visual change threshold, typing timeout, scroll timeout via the UI
• Secure API key storage — uses Electron's safeStorage for encrypted key persistence
• Usage tracking — monitors API requests, token usage, and costs

Known Issues & Limitations

• macOS ARM64 only — this release is Apple Silicon only; Intel Mac, Windows, and Linux builds are not yet available

Installation

curl -fsSL https://raw.githubusercontent.com/deusXmachina-dev/memorylane/main/install.sh | sh

This downloads the latest release and installs it to /Applications. No Gatekeeper warnings.

After launching:

1. Grant Screen Recording permission when prompted
2. Grant Accessibility permission when prompted
3. Choose how to provide an API key:
   • Subscribe (recommended) — click Subscribe to get a managed key ($20/mo via Stripe)
   • Bring Your Own Key — paste your OpenRouter API key if you already have one
4. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

https://github.com/deusXmachina-dev/memorylane/commits/v0.6.0

MemoryLane v0.5.0

MemoryLane is a macOS system tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server — giving AI assistants like Claude and Cursor memory of what you've been working on.

What's Changed

• Time tracking prompt — added a new MCP prompt that generates time reports from your screen activity, helping you understand how you spend your day
• Recent activity prompt — added a prompt for quick summaries of recent screen activity
• Apple notarization — the app is now fully notarized by Apple, eliminating Gatekeeper warnings on fresh installs (closes #5)
• MCP server restructure — split the MCP server into modular files (tools, prompts, formatting) for better maintainability
• Server instructions — embedded server instructions into the MCP server to improve AI assistant interactions
• Local install script — added a script for installing from a local build during development

Features

• One-command install — curl | sh installer that downloads, installs, and removes quarantine automatically
• Apple notarized — the app is code-signed and Apple-notarized, no Gatekeeper warnings
• Managed API key via Stripe — subscribe and start capturing in seconds, no OpenRouter account needed
• Multi-screen capture — captures screenshots from all connected displays simultaneously
• Event-driven screen capture — captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• OCR via macOS Vision — extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization — classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search — vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server — exposes search_context, browse_timeline, and get_event_details tools plus time tracking and recent activity prompts for AI assistants
• One-click integrations — register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture settings — adjust visual change threshold, typing timeout, scroll timeout via the UI
• Secure API key storage — uses Electron's safeStorage for encrypted key persistence
• Usage tracking — monitors API requests, token usage, and costs

Known Issues & Limitations

• macOS ARM64 only — this release is Apple Silicon only; Intel Mac, Windows, and Linux builds are not yet available

Installation

curl -fsSL https://raw.githubusercontent.com/deusXmachina-dev/memorylane/main/install.sh | sh

This downloads the latest release and installs it to /Applications. No Gatekeeper warnings.

After launching:

1. Grant Screen Recording permission when prompted
2. Grant Accessibility permission when prompted
3. Choose how to provide an API key:
   • Subscribe (recommended) — click Subscribe to get a managed key ($20/mo via Stripe)
   • Bring Your Own Key — paste your OpenRouter API key if you already have one
4. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

https://github.com/deusXmachina-dev/memorylane/commits/v0.5.0

MemoryLane v0.4.3

MemoryLane is a macOS system tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server — giving AI assistants like Claude and Cursor memory of what you've been working on.

What's Changed

• Stripe subscription for simpler AI setup — subscribe ($20/mo) and get a managed OpenRouter API key provisioned automatically. No need to create an OpenRouter account or manage keys yourself. Bring Your Own Key is still supported for users who prefer it.
• Cleaner UI for managed subscribers — API cost details are hidden when using a managed key, keeping the interface focused
• Updated README — added screenshots, reorganized sections, and made the subscription option more prominent

Features

• One-command install — curl | sh installer that downloads, installs, and removes quarantine automatically
• Managed API key via Stripe — subscribe and start capturing in seconds, no OpenRouter account needed
• Multi-screen capture — captures screenshots from all connected displays simultaneously
• Event-driven screen capture — captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• OCR via macOS Vision — extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization — classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search — vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server — exposes search_context, browse_timeline, and get_event_details tools for AI assistants
• One-click integrations — register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture settings — adjust visual change threshold, typing timeout, scroll timeout via the UI
• Secure API key storage — uses Electron's safeStorage for encrypted key persistence
• Usage tracking — monitors API requests, token usage, and costs

Known Issues & Limitations

• Not notarized — the app is code-signed but not Apple-notarized yet; install via the curl command above to avoid Gatekeeper warnings (#5)
• macOS ARM64 only — this release is Apple Silicon only; Intel Mac, Windows, and Linux builds are not yet available

Installation

curl -fsSL https://raw.githubusercontent.com/deusXmachina-dev/memorylane/main/install.sh | sh

This downloads the latest release and installs it to /Applications. No Gatekeeper warnings.

After launching:

1. Grant Screen Recording permission when prompted
2. Grant Accessibility permission when prompted
3. Choose how to provide an API key:
   • Subscribe (recommended) — click Subscribe to get a managed key ($20/mo via Stripe)
   • Bring Your Own Key — paste your OpenRouter API key if you already have one
4. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

https://github.com/deusXmachina-dev/memorylane/commits/v0.4.3

MemoryLane v0.4.2

MemoryLane is a macOS system tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server — giving AI assistants like Claude and Cursor memory of what you've been working on.

What's Changed

• Simplified API key setup — removed the Stripe subscription option for now; the setup screen shows only the Bring Your Own Key flow until managed billing is running reliably
• Download progress bar — the install.sh installer now shows a progress bar while downloading

Features

• One-command install — curl | sh installer that downloads, installs, and removes quarantine automatically
• Multi-screen capture — captures screenshots from all connected displays simultaneously
• Event-driven screen capture — captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• OCR via macOS Vision — extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization — classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search — vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server — exposes search_context, browse_timeline, and get_event_details tools for AI assistants
• One-click integrations — register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture settings — adjust visual change threshold, typing timeout, scroll timeout via the UI
• Secure API key storage — uses Electron's safeStorage for encrypted key persistence
• Usage tracking — monitors API requests, token usage, and costs

Known Issues & Limitations

• Not notarized — the app is code-signed but not Apple-notarized yet; install via the curl command above to avoid Gatekeeper warnings (#5)
• macOS ARM64 only — this release is Apple Silicon only; Intel Mac, Windows, and Linux builds are not yet available

Installation

curl -fsSL https://raw.githubusercontent.com/deusXmachina-dev/memorylane/main/install.sh | sh

This downloads the latest release and installs it to /Applications. No Gatekeeper warnings.

After launching:

1. Grant Screen Recording permission when prompted
2. Grant Accessibility permission when prompted
3. Enter your OpenRouter API key
4. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

https://github.com/deusXmachina-dev/memorylane/commits/v0.4.2

MemoryLane v0.4.1

MemoryLane is a macOS system tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server — giving AI assistants like Claude and Cursor memory of what you've been working on.

What's Changed

• Stripe subscription for simpler AI setup — subscribe ($20/mo) and get a managed OpenRouter API key provisioned automatically. No need to create an OpenRouter account or manage keys yourself. Bring Your Own Key is still supported for users who prefer it.
• Cleaner UI for managed subscribers — API cost details are hidden when using a managed key, keeping the interface focused
• Updated README — added screenshots, reorganized sections, and made the subscription option more prominent

Features

• One-command install — curl | sh installer that downloads, installs, and removes quarantine automatically
• Managed API key via Stripe — subscribe and start capturing in seconds, no OpenRouter account needed
• Multi-screen capture — captures screenshots from all connected displays simultaneously
• Event-driven screen capture — captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• OCR via macOS Vision — extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization — classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search — vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server — exposes search_context, browse_timeline, and get_event_details tools for AI assistants
• One-click integrations — register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture settings — adjust visual change threshold, typing timeout, scroll timeout via the UI
• Secure API key storage — uses Electron's safeStorage for encrypted key persistence
• Usage tracking — monitors API requests, token usage, and costs

Known Issues & Limitations

• Not notarized — the app is code-signed but not Apple-notarized yet; install via the curl command above to avoid Gatekeeper warnings (#5)
• macOS ARM64 only — this release is Apple Silicon only; Intel Mac, Windows, and Linux builds are not yet available

Installation

curl -fsSL https://raw.githubusercontent.com/deusXmachina-dev/memorylane/main/install.sh | sh

This downloads the latest release and installs it to /Applications. No Gatekeeper warnings.

After launching:

1. Grant Screen Recording permission when prompted
2. Grant Accessibility permission when prompted
3. Choose how to provide an API key:
   • Subscribe (recommended) — click Subscribe to get a managed key ($20/mo via Stripe)
   • Bring Your Own Key — paste your OpenRouter API key if you already have one
4. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

https://github.com/deusXmachina-dev/memorylane/commits/v0.4.1

MemoryLane v0.3.1

MemoryLane is a macOS system tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server — giving AI assistants like Claude and Cursor memory of what you've been working on.

What's Changed

• Curl-based installer — no more Gatekeeper warnings; install with a single command instead of manually approving the app in System Settings
• Subscription groundwork — device identity, managed API key provisioning, and Stripe checkout flow (not yet user-facing)
• UI polish — improved subscribe and key management sections

Features

• One-command install — curl | sh installer that downloads, installs, and removes quarantine automatically
• Multi-screen capture — captures screenshots from all connected displays simultaneously
• Event-driven screen capture — captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• OCR via macOS Vision — extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization — classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search — vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server — exposes search_context, browse_timeline, and get_event_details tools for AI assistants
• One-click integrations — register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture settings — adjust visual change threshold, typing timeout, scroll timeout via the UI
• Secure API key storage — uses Electron's safeStorage for encrypted key persistence
• Usage tracking — monitors API requests, token usage, and costs

Known Issues & Limitations

• Not notarized — the app is code-signed but not Apple-notarized yet; install via the curl command above to avoid Gatekeeper warnings (#5)
• macOS ARM64 only — this release is Apple Silicon only; Intel Mac, Windows, and Linux builds are not yet available

Installation

curl -fsSL https://raw.githubusercontent.com/deusXmachina-dev/memorylane/main/install.sh | sh

This downloads the latest release and installs it to /Applications. No Gatekeeper warnings.

After launching:

1. Grant Screen Recording permission when prompted
2. Grant Accessibility permission when prompted
3. Set your OpenRouter API key in the app window
4. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

https://github.com/deusXmachina-dev/memorylane/commits/v0.3.1

MemoryLane v0.3.0

MemoryLane is a macOS system tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server — giving AI assistants like Claude and Cursor memory of what you've been working on.

What's Changed

• Multi-screen capture — captures from all connected displays, not just the primary screen (closes #4)
• Different-size screen handling — correctly captures screens with different resolutions and dimensions

Features

• Multi-screen capture — captures screenshots from all connected displays simultaneously
• Event-driven screen capture — captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• OCR via macOS Vision — extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization — classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search — vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server — exposes search_context, browse_timeline, and get_event_details tools for AI assistants
• One-click integrations — register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture settings — adjust visual change threshold, typing timeout, scroll timeout via the UI
• Secure API key storage — uses Electron's safeStorage for encrypted key persistence
• Usage tracking — monitors API requests, token usage, and costs

Known Issues & Limitations

• Not notarized — the app is code-signed but not Apple-notarized yet; you'll need to right-click → Open on first launch to bypass Gatekeeper (#5)
• macOS ARM64 only — this release includes a macOS Apple Silicon DMG only; Intel Mac, Windows, and Linux builds are not yet available

Installation

1. Download MemoryLane-0.3.0-arm64.dmg
2. Open the DMG and drag MemoryLane to Applications
3. Double-click the app to open it — you'll see a warning that it can't be opened because Apple cannot check it for malicious software
4. Go to System Settings → Privacy & Security, scroll down, and click "Open Anyway" next to the MemoryLane message
5. Authenticate with your password or Touch ID and confirm the final dialog
6. Grant Screen Recording permission when prompted
7. Set your OpenRouter API key in the app window
8. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

https://github.com/deusXmachina-dev/memorylane/commits/v0.3.0

MemoryLane v0.2.0

MemoryLane is a macOS system tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server — giving AI assistants like Claude and Cursor memory of what you've been working on.

What's Changed

• Performance improvements — addressed resource-intensive capture and processing issues with debouncing, max capture rate limiting, and fewer screenshots (closes #3)
• Single-node-env migration — all scripts and tests now run under Electron's Node.js runtime, eliminating ABI mismatch errors
• "Add to Claude Code" button — one-click MCP server registration for Claude Code from the tray menu
• SQLite vector size fix — resolved max vector size issue for sqlite-vec
• Build and test fixes — improved stability and reliability

Features

• Event-driven screen capture — captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• OCR via macOS Vision — extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization — classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search — vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server — exposes search_context, browse_timeline, and get_event_details tools for AI assistants
• One-click integrations — register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture settings — adjust visual change threshold, typing timeout, scroll timeout via the UI
• Secure API key storage — uses Electron's safeStorage for encrypted key persistence
• Usage tracking — monitors API requests, token usage, and costs

Known Issues & Limitations

• Single display only — currently captures from one screen; multi-monitor support is not yet implemented (#4)
• Not notarized — the app is code-signed but not Apple-notarized yet; you'll need to right-click → Open on first launch to bypass Gatekeeper (#5)
• macOS ARM64 only — this release includes a macOS Apple Silicon DMG only; Intel Mac, Windows, and Linux builds are not yet available

Installation

1. Download MemoryLane-0.2.0-arm64.dmg
2. Open the DMG and drag MemoryLane to Applications
3. Double-click the app to open it — you'll see a warning that it can't be opened because Apple cannot check it for malicious software
4. Go to System Settings → Privacy & Security, scroll down, and click "Open Anyway" next to the MemoryLane message
5. Authenticate with your password or Touch ID and confirm the final dialog
6. Grant Screen Recording permission when prompted
7. Set your OpenRouter API key in the app window
8. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

https://github.com/deusXmachina-dev/memorylane/commits/v0.2.0

MemoryLane v0.1.0 — First Alpha Release

MemoryLane is a macOS system tray app that captures your screen activity, processes it with OCR and AI summarization, and makes it searchable through an MCP server — giving AI assistants like Claude and Cursor memory of what you've been working on.

Features

• Event-driven screen capture — captures screenshots based on user interactions (clicks, typing, scrolling, app switches) and visual changes (perceptual dHash comparison), not fixed intervals
• OCR via macOS Vision — extracts text from screenshots using the native Vision framework (Swift sidecar)
• AI-powered summarization — classifies activity into concise summaries using vision models via OpenRouter (Mistral Small, GPT-5 Nano, Grok-4.1 Fast, Gemini Flash Lite)
• Semantic search — vector embeddings (all-MiniLM-L6-v2) + SQLite FTS5 for full-text and semantic search over your activity history
• MCP server — exposes search_context, browse_timeline, and get_event_details tools for AI assistants
• One-click integrations — register the MCP server with Claude Desktop or Cursor from the tray menu
• Configurable capture settings — adjust visual change threshold, typing timeout, scroll timeout via the UI
• Secure API key storage — uses Electron's safeStorage for encrypted key persistence
• Usage tracking — monitors API requests, token usage, and costs

Known Issues & Limitations

• Performance — capture and processing can be resource-intensive, especially with frequent visual changes (#3)
• Single display only — currently captures from one screen; multi-monitor support is not yet implemented (#4)
• Not notarized — the app is code-signed but not Apple-notarized yet; you'll need to right-click → Open on first launch to bypass Gatekeeper (#5)
• macOS ARM64 only — this release includes a macOS Apple Silicon DMG only; Intel Mac, Windows, and Linux builds are not yet available

Installation

1. Download MemoryLane-0.1.0-arm64.dmg
2. Open the DMG and drag MemoryLane to Applications
3. Double-click the app to open it — you'll see a warning that it can't be opened because Apple cannot check it for malicious software
4. Go to System Settings → Privacy & Security, scroll down, and click "Open Anyway" next to the MemoryLane message
5. Authenticate with your password or Touch ID and confirm the final dialog
6. Grant Screen Recording permission when prompted
7. Set your OpenRouter API key in the app window
8. Optionally register the MCP server with Claude Desktop or Cursor

Full Changelog

https://github.com/deusXmachina-dev/memorylane/commits/v0.1.0