Limetry

Changelog

What's new in Limetry

Notable releases across the TypeScript SDK, CLI, evaluation server, and MCP packages.

  1. Feature

    Public website: docs, blog, examples, and pricing

    Launched limetry.com with Fumadocs-powered documentation, an engineering blog, a browsable example gallery, and community and pricing pages.

  2. Feature

    @limetry/ui published to npm

    The shared UI component library is now published publicly, with pinned dependency versions and tightened peer dependency ranges across the monorepo.

  3. Docs

    USAGE.md integration guide

    Published USAGE.md: install the SDK, evaluate policies, enforce outcomes, and query privacy-safe audit.

  4. Infra

    Continuous integration pipeline

    Added GitHub Actions running SDK tests and a smoke test across integration examples on every push.

  5. Fix

    Tighter public SDK types

    Tighter TypeScript definitions, JSDoc on the public API, and broader ESLint coverage.

  6. Feature

    AI agent skill plugin

    Added a Gemini / Antigravity agent skill so coding agents can find and call Limetry evaluate.

  7. Feature

    @limetry/mcp — MCP server

    New Model Context Protocol server that brings policy evaluation and audit tools to Claude, Cursor, and Copilot.

  8. Feature

    Integration examples

    Shipped examples for OpenAI, LangChain, AI SDK agent loops, CrewAI, AutoGen, Slack bots, ChatGPT Custom GPTs, and related agent stacks.

  9. Feature

    @limetry/cli

    New CLI with limetry setup, token, and auth subcommands for provisioning projects and credentials from the terminal.

  10. Feature

    Self-run evaluation server

    Open-source evaluation server for remote policy evaluation and audit, plus the Limetry SDK for HTTP evaluate against a Limetry server.

  11. Feature

    TypeScript SDK

    Scaffolded @limetry/sdk with the Limetry SDK, action intent helpers, and typed evaluate errors.

  12. Feature

    Action policy evaluate

    Initial release: check action intents against action policy (types, resource patterns, optional cost caps), privacy-safe audit, and signed outcome receipts.