Fixes
- Memory leak in memory:// docket broker โ cancelled tasks now properly cleaned up instead of accumulating
- Bumps pydocket to โฅ0.17.2 (contains the fix)
Unclaimed project
Are you a maintainer of fastmcp? Claim this project to take control of your public changelog and roadmap.
Changelog
๐ The fast, Pythonic way to build MCP servers and clients
Last updated about 2 months ago
This patch release fixes a fresh install bug where the packaging library was previously installed as a transitive dependency but is no longerโcausing an import error on fresh installs without dev dependencies. Also includes a pydocket version pin to avoid Redis connection noise in tests, plus backports from 3.x for $ref dereferencing in tool schemas and the task capabilities location fix.
FastMCP 3.0 rebuilds the framework around three primitives: components, providers, and transforms. Providers source components dynamicallyโfrom decorators, filesystems, OpenAPI specs, remote servers, or anywhere else. Transforms modify components as they flow to clientsโrenaming, namespacing, filtering, securing. The features that required specialized subsystems in v2 now compose naturally from th...
This patch release fixes an HTTP transport timeout bug where connections were defaulting to 5 seconds instead of respecting MCP's 30-second default, causing premature timeouts for slower operations. Also includes OAuth token storage fixes, Redis key isolation for ACL compliance, and improved ContextVar propagation for ASGI-mounted servers. Plus, the CLI will now nudge you when updates are availabl...
FastMCP 2.14.2 brings a wave of community contributions safely into the 2.x line. A variety of important fixes backported from 3.0 work improve OpenAPI 3.1 compatibility, MCP spec compliance for output schemas and elicitation, and correct a subtle base_url fallback issue. The CLI now gently reminds you that FastMCP 3.0 is on the horizon.
FastMCP 2.14.1 adds support for sampling with tools (SEP-1577). This exciting new feature lets servers pass tools to ctx.sample(), enabling agentic workflows where the server borrows the client's LLM and controls tool execution automatically. Pass any callable as a tool and FastMCP handles the loop: calling the LLM, executing tools, and feeding results back until a final response is produced. Fo...
FastMCP 2.14 begins adopting the MCP 2025-11-25 specification, headlined by protocol-native background tasks that let long-running operations report progress without blocking clients. This release also graduates the OpenAPI parser to standard, adds first-class support for several new spec features, and removes deprecated APIs accumulated across the 2.x series.
MCP SDK 1.23 introduced some changes related to the 11/25/25 MCP protocol update that break some patches/workarounds that FastMCP had implemented previously. In particular, OAuth changes in the new protocol changed some implementation details that FastMCP patched; as such 1.23 is not necessarily a breaking SDK change but it is "breaking" for certain FastMCP behaviors.
As a precaution, this rele...
FastMCP 2.13.2 polishes the authentication stack with fixes for token refresh, scope handling, and multi-instance deployments. Discord joins the growing roster of built-in OAuth providers, Azure and Google token handling gets more reliable, and proxy classes now properly forward icons and titles. This release also adds CSP customization for consent screens and fixes an edge case where $defs coul...
FastMCP 2.13.1 introduces meta parameter support for ToolResult (#2283), letting tools return metadata alongside results to enable new use cases such as OpenAI's Apps SDK. It also supports client-sent meta (#2206) as well as improved OAuth capabilities and custom token verifiers (including the new DebugTokenVerifier) and an OCI authentication provider. A large list of enhancements and bugfixes...
AutoGPT is the vision of accessible AI for everyone, to use and to build on. Our mission is to provide the tools, so that you can focus on what matters.
Stable Diffusion web UI
๐ค Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.
A feature-rich command-line audio/video downloader