Skip to content

distillation-labs/agentyc

v0.4.1 Feature

This release adds 5 notable features for engineering teams evaluating rollout.

✓ No known CVEs patched
Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

Summary

AI summary

Shadow DOM support, deterministic dialog handling, and deadlock fix improve reliability.

Changes in this release

Feature Medium

`browser_get_state` now pierces open Shadow DOM roots, making shadow inputs/buttons reachable as stable refs.

`browser_get_state` now pierces open Shadow DOM roots, making shadow inputs/buttons reachable as stable refs.

Source: llm_adapter@2026-06-14

Confidence: high

Feature Medium

Native JavaScript dialogs are auto-handled (default accept) with `browser_handle_dialog` controlling policy.

Native JavaScript dialogs are auto-handled (default accept) with `browser_handle_dialog` controlling policy.

Source: llm_adapter@2026-06-14

Confidence: high

Feature Medium

Element resolution retries once on transient layout failures (stale‑ref auto‑retry).

Element resolution retries once on transient layout failures (stale‑ref auto‑retry).

Source: llm_adapter@2026-06-14

Confidence: high

Feature Medium

15 new observability tools available via `--extended` flag (`AGENTYC_EXTENDED=1`).

15 new observability tools available via `--extended` flag (`AGENTYC_EXTENDED=1`).

Source: llm_adapter@2026-06-14

Confidence: high

Feature Medium

`browser_get_html` and `browser_extract_content` are token‑bounded to prevent blowing agent context windows.

`browser_get_html` and `browser_extract_content` are token‑bounded to prevent blowing agent context windows.

Source: llm_adapter@2026-06-14

Confidence: high

Feature Medium

'min' mode is now the default for `get_state`, providing smallest reliable payload.

'min' mode is now the default for `get_state`, providing smallest reliable payload.

Source: llm_adapter@2026-06-14

Confidence: high

Bugfix Medium

All `cdp()` helpers release the state lock before awaiting, fixing deadlock on click→dialog flows.

All `cdp()` helpers release the state lock before awaiting, fixing deadlock on click→dialog flows.

Source: llm_adapter@2026-06-14

Confidence: high

Full changelog

What's new in v0.4.1

  • Shadow DOM supportbrowser_get_state pierces open shadow roots; shadow inputs and buttons are now reachable as stable refs.
  • Deterministic dialog handling — native JS dialogs are auto-handled (default: accept). browser_handle_dialog sets the policy for subsequent dialogs.
  • Stale-ref auto-retry — element resolution retries once on transient layout failures.
  • Extended observability profile — 15 new tools available via --extended (AGENTYC_EXTENDED=1): console logs, network log, mocks, conditions, request replay, debug bundle, downloads, trace.
  • Token-bounded outputsbrowser_get_html and browser_extract_content are now bounded to avoid blowing agent context windows.
  • min is the default get_state mode — smallest reliable payload for agents by default.
  • Deadlock fix — all cdp() helpers now release the state lock before awaiting, preventing any click→dialog hangs.

Install

Download a prebuilt binary below (no Rust toolchain required):

  • agentyc-aarch64-apple-darwin.tar.gz — macOS Apple Silicon
  • agentyc-x86_64-apple-darwin.tar.gz — macOS Intel
  • agentyc-x86_64-unknown-linux-gnu.tar.gz — Linux x86_64
  • agentyc-x86_64-pc-windows-msvc.zip — Windows x86_64

Or build from source:

cargo install --git https://github.com/distillation-labs/agentyc agentyc

MCP config (Claude Code / Cursor)

{ "mcpServers": { "agentyc": { "command": "agentyc", "args": ["mcp"] } } }

Add "--extended" to args to unlock the 15 observability tools.

See CHANGELOG.md for full details.

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

Share this release

Track distillation-labs/agentyc

Get notified when new releases ship.

Sign up free

About distillation-labs/agentyc

All releases →

Beta — feedback welcome: [email protected]