base on Bridge between Ollama and MCP servers, enabling local LLMs to use Model Context Protocol tools # MCP-LLM Bridge A TypeScript implementation that connects local LLMs (via Ollama) to Model Context Protocol (MCP) servers. This bridge allows open-source models to use the same tools and capabilities as Claude, enabling powerful local AI assistants. ## Overview This project bridges local Large Language Models with MCP servers that provide various capabilities like: - Filesystem operations - Brave web search - GitHub interactions - Google Drive & Gmail integration - Memory/storage - Image generation with Flux The bridge translates between the LLM's outputs and the MCP's JSON-RPC protocol, allowing any Ollama-compatible model to use these tools just like Claude does. ## Current Setup - **LLM**: Using Qwen 2.5 7B (qwen2.5-coder:7b-instruct) through Ollama - **MCPs**: - Filesystem operations (`@modelcontextprotocol/server-filesystem`) - Brave Search (`@modelcontextprotocol/server-brave-search`) - GitHub (`@modelcontextprotocol/server-github`) - Memory (`@modelcontextprotocol/server-memory`) - Flux image generation (`@patruff/server-flux`) - Gmail & Drive (`@patruff/server-gmail-drive`) ## Architecture - **Bridge**: Core component that manages tool registration and execution - **LLM Client**: Handles Ollama interactions and formats tool calls - **MCP Client**: Manages MCP server connections and JSON-RPC communication - **Tool Router**: Routes requests to appropriate MCP based on tool type ### Key Features - Multi-MCP support with dynamic tool routing - Structured output validation for tool calls - Automatic tool detection from user prompts - Robust process management for Ollama - Detailed logging and error handling ## Setup 1. Install Ollama and required model: ```bash ollama pull qwen2.5-coder:7b-instruct ``` 2. Install MCP servers: ```bash npm install -g @modelcontextprotocol/server-filesystem npm install -g @modelcontextprotocol/server-brave-search npm install -g @modelcontextprotocol/server-github npm install -g @modelcontextprotocol/server-memory npm install -g @patruff/server-flux npm install -g @patruff/server-gmail-drive ``` 3. Configure credentials: - Set `BRAVE_API_KEY` for Brave Search - Set `GITHUB_PERSONAL_ACCESS_TOKEN` for GitHub - Set `REPLICATE_API_TOKEN` for Flux - Run Gmail/Drive MCP auth: `node path/to/gmail-drive/index.js auth` - For example node C:\Users\patru\AppData\Roaming\npm\node_modules\@patruff\server-gmail-drive\dist\index.js auth ## Configuration The bridge is configured through `bridge_config.json`: - MCP server definitions - LLM settings (model, temperature, etc.) - Tool permissions and paths Example: ```json { "mcpServers": { "filesystem": { "command": "node", "args": ["path/to/server-filesystem/dist/index.js"], "allowedDirectory": "workspace/path" }, // ... other MCP configurations }, "llm": { "model": "qwen2.5-coder:7b-instruct", "baseUrl": "http://localhost:11434" } } ``` ## Usage 1. Start the bridge: ```bash npm run start ``` 2. Available commands: - `list-tools`: Show available tools - Regular text: Send prompts to the LLM - `quit`: Exit the program Example interactions: ``` > Search the web for "latest TypeScript features" [Uses Brave Search MCP to find results] > Create a new folder called "project-docs" [Uses Filesystem MCP to create directory] > Send an email to [email protected] [Uses Gmail MCP to compose and send email] ``` ## Technical Details ### Tool Detection The bridge includes smart tool detection based on user input: - Email operations: Detected by email addresses and keywords - Drive operations: Detected by file/folder keywords - Search operations: Contextually routed to appropriate search tool ### Response Processing Responses are processed through multiple stages: 1. LLM generates structured tool calls 2. Bridge validates and routes to appropriate MCP 3. MCP executes operation and returns result 4. Bridge formats response for user ## Extended Capabilities This bridge effectively brings Claude's tool capabilities to local models: - Filesystem manipulation - Web search and research - Email and document management - Code and GitHub interactions - Image generation - Persistent memory All while running completely locally with open-source models. ## Future Improvements - Add support for more MCPs - Implement parallel tool execution - Add streaming responses - Enhance error recovery - Add conversation memory - Support more Ollama models ## Related Projects This bridge integrates with the broader Claude ecosystem: - Model Context Protocol (MCP) - Claude Desktop Configuration - Ollama Project - Various MCP server implementations The result is a powerful local AI assistant that can match many of Claude's capabilities while running entirely on your own hardware. ", Assign "at most 3 tags" to the expected json: {"id":"13175","tags":[]} "only from the tags list I provide: [{"id":77,"name":"3d"},{"id":89,"name":"agent"},{"id":17,"name":"ai"},{"id":54,"name":"algorithm"},{"id":24,"name":"api"},{"id":44,"name":"authentication"},{"id":3,"name":"aws"},{"id":27,"name":"backend"},{"id":60,"name":"benchmark"},{"id":72,"name":"best-practices"},{"id":39,"name":"bitcoin"},{"id":37,"name":"blockchain"},{"id":1,"name":"blog"},{"id":45,"name":"bundler"},{"id":58,"name":"cache"},{"id":21,"name":"chat"},{"id":49,"name":"cicd"},{"id":4,"name":"cli"},{"id":64,"name":"cloud-native"},{"id":48,"name":"cms"},{"id":61,"name":"compiler"},{"id":68,"name":"containerization"},{"id":92,"name":"crm"},{"id":34,"name":"data"},{"id":47,"name":"database"},{"id":8,"name":"declarative-gui "},{"id":9,"name":"deploy-tool"},{"id":53,"name":"desktop-app"},{"id":6,"name":"dev-exp-lib"},{"id":59,"name":"dev-tool"},{"id":13,"name":"ecommerce"},{"id":26,"name":"editor"},{"id":66,"name":"emulator"},{"id":62,"name":"filesystem"},{"id":80,"name":"finance"},{"id":15,"name":"firmware"},{"id":73,"name":"for-fun"},{"id":2,"name":"framework"},{"id":11,"name":"frontend"},{"id":22,"name":"game"},{"id":81,"name":"game-engine "},{"id":23,"name":"graphql"},{"id":84,"name":"gui"},{"id":91,"name":"http"},{"id":5,"name":"http-client"},{"id":51,"name":"iac"},{"id":30,"name":"ide"},{"id":78,"name":"iot"},{"id":40,"name":"json"},{"id":83,"name":"julian"},{"id":38,"name":"k8s"},{"id":31,"name":"language"},{"id":10,"name":"learning-resource"},{"id":33,"name":"lib"},{"id":41,"name":"linter"},{"id":28,"name":"lms"},{"id":16,"name":"logging"},{"id":76,"name":"low-code"},{"id":90,"name":"message-queue"},{"id":42,"name":"mobile-app"},{"id":18,"name":"monitoring"},{"id":36,"name":"networking"},{"id":7,"name":"node-version"},{"id":55,"name":"nosql"},{"id":57,"name":"observability"},{"id":46,"name":"orm"},{"id":52,"name":"os"},{"id":14,"name":"parser"},{"id":74,"name":"react"},{"id":82,"name":"real-time"},{"id":56,"name":"robot"},{"id":65,"name":"runtime"},{"id":32,"name":"sdk"},{"id":71,"name":"search"},{"id":63,"name":"secrets"},{"id":25,"name":"security"},{"id":85,"name":"server"},{"id":86,"name":"serverless"},{"id":70,"name":"storage"},{"id":75,"name":"system-design"},{"id":79,"name":"terminal"},{"id":29,"name":"testing"},{"id":12,"name":"ui"},{"id":50,"name":"ux"},{"id":88,"name":"video"},{"id":20,"name":"web-app"},{"id":35,"name":"web-server"},{"id":43,"name":"webassembly"},{"id":69,"name":"workflow"},{"id":87,"name":"yaml"}]" returns me the "expected json"