home / mcp / mcp bridge mcp server

MCP Bridge MCP Server

Provides local MCP endpoints for filesystem, web search, GitHub, memory, image generation, and Gmail/Drive access

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "patruff-ollama-mcp-bridge": {
      "command": "node",
      "args": [
        "path/to/server-filesystem/dist/index.js"
      ],
      "env": {
        "BRAVE_API_KEY": "YOUR_BRAVE_API_KEY",
        "REPLICATE_API_TOKEN": "YOUR_REPLICATE_TOKEN",
        "GITHUB_PERSONAL_ACCESS_TOKEN": "YOUR_GITHUB_TOKEN"
      }
    }
  }
}

You run a local AI assistant that can access powerful MCP tools through a bridge. This MCP Server embraces multiple locally hosted MCP endpoints and connects them to your local LLM, enabling capabilities like filesystem operations, web search, GitHub interactions, email and drive management, memory, and image generation. It executes tool calls from your model and returns structured results for seamless, private AI workflows.

How to use

Start by launching the bridge with its start command. Once running, you can chat with your local LLM and ask it to perform tasks that leverage the connected MCP servers. The bridge will detect when your prompt implies a tool use (for example filesystem actions, web searches, or email tasks) and route the request to the appropriate MCP server. You will see the tool being used in responses, followed by the operation result and any follow-up actions required.

Common usage patterns include: creating directories or files, performing web searches for research, managing GitHub repositories, drafting and sending emails, and generating images. The tool routing is automatic, so you don’t have to manually choose a tool. You can still list available tools to understand what the system can do, or simply prompt in natural language and let the bridge call the right MCP.

If you want to inspect the tools available at any time, use the list-tools command provided by the bridge. You can then continue with your regular prompts to perform actions with the corresponding MCPs.

How to install

Prerequisites you need before installing: Node.js installed on your system and access to a terminal or command prompt. You will also need Ollama installed to run local LLMs.

Install Ollama and pull a local model suitable for your setup. For example, pull a Qwen model compatible with Ollama.

Install the MCP servers that your bridge will connect to. Execute these commands to install the MCP endpoints used by the bridge.

Install the GAP to run CLI commands for the Gmail/Drive integration as needed for authentication and setup.

Set up credentials for the MCP services you plan to use. Prepare keys and tokens for Brave Search, GitHub, and image generation services as required. These credentials will be referenced by the bridge during tool calls.

Create the bridge configuration file, following the example layout to specify the MCP server definitions and LLM settings. This file determines which MCPs are available and how the LLM should connect to them.

Start the bridge and begin your session. The bridge will load the MCPs, initialize the LLM interface, and enable tool routing for your prompts.

Configuration and security

Your configuration includes the MCP server definitions, LLM parameters (model name, temperature, etc.), and tool permissions. Ensure you provide secure access to credentials and keep them out of version control. Use environment variables to manage sensitive keys and tokens.

Common credentials you will configure include Brave Search API key, GitHub personal access token, and a token for image generation or replication services. Store these securely and reference them in your bridge configuration as shown in the example env entries.

If you run into issues, review the bridge logs for tool routing messages and MCP server responses. The bridge provides clear error handling and logging to help you diagnose failures in tool execution or MCP communication.

Examples of supported capabilities

Filesystem operations allow you to manage files and directories on your workspace from prompts.

Brave Search lets you perform web research directly from your prompt and returns structured results.

GitHub interactions enable repository exploration, issue management, and code retrieval.

Memory and storage capabilities let you persist context across sessions for more coherent interactions.

Flux image generation supports creating images through the integrated tool.

Gmail and Drive functionality integrates email and cloud storage tasks, such as composing messages and managing files.

Notes and tips

The bridge is designed for local operation with open-source models and supports multiple MCP servers simultaneously. Ensure each MCP is properly configured and reachable by the bridge. Start with a minimal setup and gradually enable additional MCPs as you validate the workflow.

Available tools

filesystem

Filesystem operations: create, delete, move, list directories and files within the workspace.

brave_search

Web search tool using Brave Search to fetch results and citations for research tasks.

github

GitHub integration for repository interactions, issue management, and code retrieval.

memory

Persistent memory/storage to maintain context across sessions and prompts.

flux

Flux image generation for creating images programmatically.

gmail_drive

Gmail and Google Drive integration for composing emails and managing files.