This release adds 3 notable features for engineering teams evaluating rollout.
✓ No known CVEs patched in this version
Topics
+2 more
Summary
AI summaryUpdates human-in-the-loop, https://github.com/eidentic/eidentic/blob/main/docs/BENCHMARKS.md, and https://github.com/eidentic/eidentic/blob/main/STABILITY.md across a mixed release.
Full changelog
Initial release of Eidentic — agents that remember.
An open-source TypeScript SDK for building AI agents with self-improving memory and production fundamentals built in.
Install: npm install eidentic · Scaffold: npm create eidentic@latest
32 packages on npm under eidentic and @eidentic/*. Highlights:
- Self-improving 4-tier memory: self-editing blocks, lexical+vector recall, temporal knowledge graph, sleep-time consolidation — plus governance: state-transition timelines, corroboration tiers, consent manifests, portable export, identity merge
- Durable execution: checkpoint/resume, workflows with suspend/resume (human-in-the-loop), per-step retry, versioned runs
- Production server: auth, pre-auth rate limiting, quotas, multi-tenant isolation, HMAC-signed webhooks, graceful drain
- Sandboxed tools (files, bash, web, browser), MCP host+server with governance, A2A
- Enforced cost ceilings, model routing/fallback/caching, prompt versioning with canary
- Eval harness with CI baselines — and the first official LoCoMo results: memory mode beats the full-context baseline on temporal reasoning (+12.1pp) and adversarial robustness (+15.7pp) at 95.3% fewer tokens per query
- React hooks, Next.js handler, Studio dev UI, CLI
- Runs on Node, Bun, Deno, and the edge. Apache-2.0.
See STABILITY for the versioning contract.
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
About Eidentic
All releases →Related context
Related tools
Beta — feedback welcome: [email protected]