DISCOVER THE FUTURE OF AI AGENTS

RMCP (Rust Model Context Protocol SDK)

Added May 8, 2026
Agent & Tooling
Open Source
RustModel Context ProtocolAI AgentsSDKCLIAgent & ToolingModel & Inference FrameworkDeveloper Tools & CodingProtocol, API & Integration

The official Rust SDK for the Model Context Protocol, providing both client and server implementations with full support for Tools, Resources, Prompts, Sampling, OAuth, and more.

RMCP is the official Rust implementation of the Model Context Protocol specification, published as the rmcp crate on crates.io at version 1.6.0. Hosted under the official MCP GitHub organization modelcontextprotocol, it is built on the tokio async runtime and licensed under Apache-2.0.

The SDK provides both Server and Client capabilities. The Server side supports declarative definition of Tools, Resources, and Prompts via procedural macros such as #[tool_router] and #[tool], with automatic JSON Schema parameter validation and route registration. The Client side connects to external MCP servers through TokioChildProcess or HTTP transports. Both are launched uniformly via ServiceExt::serve(transport).

In terms of protocol coverage, RMCP implements the core capabilities of the MCP 2025-11-25 specification, including Tools, Resources (with Resource Templates), Prompts, Sampling (server-initiated LLM completion requests), Roots, Logging (8 levels), Completions, Notifications (progress/cancellation/lifecycle/list changes), Subscriptions (resource change push), Elicitation (structured user input requests), and full OAuth 2.0 authorization flow (with client-credentials-JWT extension).

The transport layer is flexibly composed through feature flags: stdio (enabled by default with server), child process, Streamable HTTP Client/Server (with reqwest and Unix Socket backends), and Tower Service integration. The project uses a workspace structure with approximately 21k lines of core Rust code, a conformance test suite, and currently achieves SEP-1730 Tier 1 compliance at Server 87.5% and Client 80.0%.

The downstream ecosystem spans AI Agent frameworks (goose), GraphQL tooling (apollo-mcp-server), container management (containerd-mcp-server), editor integration (nvim-mcp), and more. The extension project rmcp-openapi automatically converts OpenAPI endpoints to MCP tools. First released on 2025-03-16, the project reached 1.0.0 stable on 2026-03-03 after rapid 0.x iteration, with automated testing and release via GitHub Actions CI and release-plz.

Related Projects

View All

STAY UPDATED

Get the latest AI tools and trends delivered straight to your inbox. No spam, just intelligence.