Ollama MCP Bridge

🚀 MCP-LLM Bridge

This project, implemented in TypeScript, aims to connect local large language models (via Ollama) to the Model Context Protocol (MCP) server. This bridge enables open-source models to use the same tools and features as Claude, thereby enabling a powerful local AI assistant.

🚀 Quick Start

Installation and Configuration

1. Install Ollama and the required model:

ollama pull qwen2.5-coder:7b-instruct

2. Install MCP servers:

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.
   • Configure Gmail credentials using the following command: cp $HOME/.config/Google/credentials.json ~/.config/gmail-credential

Configuration File

Create or edit the configuration file located at $HOME/.config/mcp-llm/config.json with the following example content:

{
  "ollama": {
    "host": "http://localhost:11434"
  },
  "gmail": {
    "enable": false,
    "credential_path": "~/.config/gmail-credential"
  }
}

Usage

1. Start the bridge:

npm run start

2. Example of sending requests:
   • Send a simple message: curl -X POST http://localhost:3000/api/v1/message -H "Content-Type: application/json" -d '{"message": "Hello, world!"}'
   • Send a message with tool invocation: curl -X POST http://localhost:3000/api/v1/message -H "Content-Type: application/json" -d '{"message": "Tell me a joke. Please use the joke-teller tool."}'

✨ Features

Overview

This project connects local large language models to MCP servers, which offer various features, such as:

• File system operations
• Brave web search
• GitHub interaction
• Google Drive and Gmail integration
• Memory/storage management
• Image generation using Flux

The bridge converts between the LLM's output and the MCP's JSON-RPC protocol, allowing any Ollama-compatible model to use these tools like Claude.

Current Configuration

Architecture

• Bridge: The core component responsible for tool registration and execution management.
• LLM Client: Handles Ollama interactions and formats tool calls.
• MCP Client: Manages MCP server connections and JSON-RPC communication.
• Tool Router: Routes requests to the appropriate MCP based on the tool type.

Key Features

• Support for multiple MCPs and dynamic tool routing.
• Structured output validation for tool calls.
• Automatic tool detection in user prompts.
• Powerful Ollama process management.
• Detailed logging and error handling.

🔧 Technical Details

• The bridge supports multi-threaded processing with a maximum of 100 connections.
• The request timeout is 30 seconds, and the number of retries is 3.
• It supports file attachment uploads and downloads.

📚 Documentation

Future Improvement Plans

• Implement support for more MCP servers.
• Add parallel tool execution functionality.
• Support streaming response output.
• Improve error recovery capabilities.
• Add conversation memory functionality.
• Expand support for more Ollama models.

Related Projects

This project 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 capable of many Claude-like features while running entirely on your own hardware.