MCP-LLM Bridge

Example Use Cases

• 1

  Perform filesystem operations on a local workspace (create folders, read/write files) through an MCP-backed toolset.

• 2

  Search the web and retrieve up-to-date search results using Brave Search integrated as an MCP.

• 3

  Compose and send emails, manage Google Drive files, interact with GitHub, or generate images via Flux while running entirely on local infrastructure.

Description

This repository implements a bridge that translates LLM outputs into MCP JSON-RPC calls, enabling local models to interact with a variety of MCP servers (filesystem, Brave Search, GitHub, Gmail/Drive, Flux image generation, memory, etc.). It manages tool registration, structured output validation, and dynamic routing to multiple MCP servers while handling Ollama process management and logging. The bridge is configurable via bridge_config.json and aims to provide Claude-like tool capabilities to locally hosted models.