mcp-framework

The MCP Framework Rust implementation provides a production-ready environment for building AI agents that communicate with MCP servers. It includes an MCP client/server architecture, multi-LLM integration (Anthropic/Claude and OpenAI), and a Web Inspector for debugging and testing tools. This server lets you register custom tools that agents can invoke via MCP protocol, manage conversations, and run agents that orchestrate multiple LLMs and tools in parallel. To get started, clone the repository, install Rust, and run the provided examples to see how a single tool server, a tool-rich server with an inspector, and browser automation flows work with both Claude and OpenAI backends.

Prerequisites:

• Rust toolchain (Rust 1.70+)
• Optional OpenAI or Anthropic API keys for demonstrations

1. Install Rust (if not already installed):

curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh

2. Clone the repository and navigate into it:

git clone https://github.com/koki7o/mcp-framework
cd mcp-framework

3. Set up environment variables (optional but recommended):

# Copy example env and edit keys as needed
cp .env.example .env

4. Build and run the server (minimal):

# Run the default server (one tool)
cargo run

5. Run a server with tools (examples) or with the inspector UI:

# Server with 8 tools and Inspector UI
cargo run --example server_with_tools

6. If you want to expose other tool or start with a custom server, follow the server example in the docs to implement ToolHandler and register tools, then run the binary with cargo as shown above.

Notes and tips:

• The MCP setup expects optional API keys for Claude/Anthropic or OpenAI; set ANTHROPIC_API_KEY and/or OPENAI_API_KEY in a .env file as recommended.
• The server supports an Inspector UI accessible at the provided local URL when using the examples that enable tooling and UI.
• If you modify environment variables, restart the server to pick up changes.
• Default commands assume you are running from the project root; use cargo run with appropriate flags if you launch from a different directory.
• This Rust implementation emphasizes zero-cost abstractions, async runtimes (Tokio), and safe concurrency; ensure you use compatible Rust nightly if required by certain features in examples.