MCP.so
Sign In
Servers

MCP-LLM Bridge

@patruff

Bridge between Ollama and MCP servers, enabling local LLMs to use Model Context Protocol tools

Overview

What is MCP-LLM Bridge?

A TypeScript implementation that connects local LLMs (via Ollama) to Model Context Protocol (MCP) servers. It translates between an LLM’s outputs and MCP’s JSON‑RPC protocol, letting open‑source models use the same tools and capabilities as Claude while running entirely locally.

How to use MCP-LLM Bridge?

Install Ollama, pull a model (e.g., ollama pull qwen2.5-coder:7b-instruct), and globally install the required MCP servers via npm. Set API keys/tokens in your environment. Create a bridge_config.json for MCP server definitions and LLM settings. Start the bridge with npm run start. Available commands: list‑tools (show available tools), regular text (send prompts), and quit.

Key features of MCP-LLM Bridge

  • 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

Use cases of MCP-LLM Bridge

  • Search the web for recent information (Brave Search MCP)
  • Create, read, or modify files and folders (Filesystem MCP)
  • Send emails and manage documents (Gmail & Drive MCP)
  • Interact with GitHub repositories (GitHub MCP)
  • Generate images via Flux (Flux MCP)

FAQ from MCP-LLM Bridge

Which LLM models does it support?

The bridge works with any Ollama-compatible model. The current setup uses Qwen 2.5 7B (qwen2.5-coder:7b-instruct), and future improvements plan to support more Ollama models.

What MCP servers are included?

The README lists six: Filesystem, Brave Search, GitHub, Memory, Flux image generation, and Gmail & Drive.

How does tool detection work?

The bridge automatically detects tools based on user input—for example, email addresses trigger Gmail operations, file/folder keywords trigger Drive operations, and search terms are routed to Brave Search.

Where does data live?

All processing runs completely locally with open‑source models; no cloud data transfer is mentioned.

What credentials are required?

Brave Search requires BRAVE_API_KEY, GitHub requires GITHUB_PERSONAL_ACCESS_TOKEN, Flux requires REPLICATE_API_TOKEN, and the Gmail/Drive MCP must be authenticated via node path/to/gmail-drive/index.js auth.

More from AI & Agents