mcp-rig

# MCP-Rig MCP-Rig is a Rust crate that provides seamless integration between the Model Context Protocol (MCP) and Rig's LLM agent framework. This crate allows you to expose MCP tools to Rig agents, creating a bridge between these two powerful systems. ## Overview The integration works by adapting MCP tools to implement Rig's `Tool` and `ToolEmbedding` traits, making them directly usable within Rig's agent framework. This allows you to: - Expose existing MCP tools to LLM agents - Use tools with different transport mechanisms (stdio, SSE) - Enable semantic retrieval of tools based on natural language queries - Manage multiple MCP clients in a single application ## Features - **MCP Client Management**: Easily create and manage multiple MCP clients - **Tool Adaptation**: Automatic adaptation of MCP tools to Rig's Tool interface - **Transport Options**: Support for both stdio and SSE-based transports - **Error Handling**: Comprehensive error handling for MCP and Rig operations - **RAG Compatibility**: Tools can be used with Retrieval-Augmented Generation ## Installation Add this to your `Cargo.toml`: ```toml [dependencies] mcp-rig = "0.1.0" rig-core = "0.9.1" tokio = { version = "1.32", features = ["full"] } ``` ## Quick Start Here's a simple example of how to use MCP-Rig to create a filesystem tool agent: ```rust use mcp_client::client::ClientInfo; use mcp_rig::{setup_rig_with_mcp, McpConnectionManager}; use rig::providers::openai::Client as RigClient; use std::collections::HashMap; use std::env; use std::time::Duration; #[tokio::main] async fn main() -> Result<(), Box<dyn std::error::Error>> { // Create the MCP connection manager let mut connection_manager = McpConnectionManager::with_timeout(Duration::from_secs(30)); // Get current directory for filesystem access let current_dir = env::current_dir()?; let current_dir_str = current_dir.to_string_lossy().to_string(); // Add a filesystem client connection_manager .add_stdio_client( "filesystem-client".to_string(), "npx", vec![ "-y".to_string(), "@modelcontextprotocol/server-filesystem".to_string(), current_dir_str, ], HashMap::new(), ClientInfo { name: "rig-integration-filesystem".to_string(), version: "1.0.0".to_string(), }, ) .await?; // Initialize the Rig client let rig_client = RigClient::from_env(); // Create the agent builder and model let agent_builder = rig_client.agent("gpt-4o").preamble( "You are a helpful assistant with access to filesystem tools." ); let model = rig_client.completion_model("o3-mini"); // Get the filesystem client let filesystem_client = connection_manager .get_client("filesystem-client") .ok_or("Filesystem client not found")?; // Set up the Rig agent with MCP tools let agent = setup_rig_with_mcp(filesystem_client, agent_builder, model).await?; // Use cli_chatbot to handle the interaction rig::cli_chatbot::cli_chatbot(agent).await?; Ok(()) } ``` ## Advanced Usage For more advanced usage, see the `bin/advanced_filesystem_example.rs` example, which demonstrates: - Setting up a filesystem MCP client - Configuring a Rig agent with MCP tools - Using the agent to interact with the filesystem - File creation and manipulation through the agent ## MCP Connection Manager The `McpConnectionManager` provides a simplified interface for working with multiple MCP clients: ```rust // Create a connection manager let mut manager = McpConnectionManager::new(); // Add a Git client using stdio transport manager.add_stdio_client( "git-client".to_string(), "uvx", vec!["mcp-server-git".to_string()], HashMap::new(), ClientInfo { name: "integration-git".to_string(), version: "1.0.0".to_string(), }, ).await?; // Add an Echo client using SSE transport manager.add_sse_client( "echo-client".to_string(), "http://localhost:8000/sse", HashMap::new(), ClientInfo { name: "integration-echo".to_string(), version: "1.0.0".to_string(), }, ).await?; // Retrieve clients by ID let git_client = manager.get_client("git-client"); let echo_client = manager.get_client("echo-client"); ``` ## License This project is licensed under the MIT License - see the LICENSE file for details.