mcp-sdk
Minimalistic Rust Implementation Of Model Context Protocol from Anthropic
claude mcp add --transport stdio antigmalabs-mcp-sdk cargo run --package mcp-sdk \ --env RUST_LOG="info"
How to use
This MCP server is a minimal Rust implementation of the Model Context Protocol (MCP). It focuses on providing a simple, easily understandable core for MCP tooling, including a Tools capability that demonstrates how to expose tool definitions, input schemas, and execution logic. The included examples illustrate how to define a CreateEntitiesTool and how the server can register handlers for tool invocation and resource listing. You can use this server as a learning reference or as a lightweight building block for experiments with MCP-based tooling and Claude Desktop-style clients.
To use the server, run it with the provided configuration. The server exposes tools that accept structured inputs (for example, an entities array for creation) and return results that can be consumed by MCP clients. Tools are defined with metadata (name and description), an input_schema that describes expected arguments, and a call method that processes input and returns a response. The server demonstrates basic protocol mechanics, including request handling and text-based responses, suitable for early-stage MCP experiments and integration with CLI or GUI clients.
How to install
Prerequisites:
- A system with Rust toolchain installed (rustup, cargo)
- Git installed for cloning repositories (optional if you already have the code locally)
Step 1: Install Rust (if not already installed)
- Visit https://rustup.rs/ and follow the instructions, or run:
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
Step 2: Clone the repository (or ensure you have the project files locally)
git clone https://github.com/modelcontextprotocol/mcp-sdk
cd mcp-sdk
Step 3: Build and run the MCP server
cargo build --release
# Or run directly via the package using cargo run
cargo run --package mcp-sdk
Step 4: Run with the configured command (if you’re using a separate environment setup)
- Use the mcp_config provided in this documentation to start the server in your environment.
Step 5: Verify installation
- Check logs for startup messages indicating the MCP server is ready and listening for tool requests.
Additional notes
Tips and common considerations:
- The project is a minimal, learning-oriented MCP implementation. More features may be missing compared to full SDKs; use it as a reference for core MCP concepts.
- Environment variable suggestion: set RUST_LOG to control verbosity (e.g., info, debug).
- If you plan to integrate with Claude Desktop or other clients, implement or adapt input_schema structures to match your tool definitions.
- The repository demonstrates a Tools capability with a sample CreateEntitiesTool; you can extend it by implementing additional tools and corresponding input schemas.
- For production experiments, you may want to containerize the server or adapt the cargo run invocation to your deployment workflow.
Related MCP Servers
goose
an open source, extensible AI agent that goes beyond code suggestions - install, execute, edit, and test with any LLM
cunzhi
告别AI提前终止烦恼,助力AI更加持久
probe
AI-friendly semantic code search engine for large codebases. Combines ripgrep speed with tree-sitter AST parsing. Powers AI coding assistants with precise, context-aware code understanding.
mcp-center
A centralized platform for managing and connecting MCP servers. MCP Center provides a high-performance proxy service that enables seamless communication between MCP clients and multiple MCP servers.
backlog -rust
MCP server for Backlog, project management service.
perplexity-web-api
🔍 Perplexity AI MCP without API key