MCP Bridge acts as a smart router and cache layer that sits in front of multiple MCP servers. It presents a unified interface with eight meta-tools and lazy schema loading, so your AI agents can discover, query, and execute actions across backend MCPs without loading every tool schema into context. Tools include listing available servers, listing tool names, fetching full tool schemas on demand, invoking tools with automatic result compaction, and retrieving compacted results. This keeps context lean while still giving you access to rich functionality across your MCP ecosystem. Use the provided commands to list servers, fetch schemas for specific tools, and call tools through the bridge; if results are large, they will be compacted and stored externally with a lightweight reference returned for retrieval when needed.

Prerequisites:

• Node.js and npm installed on your system
• Access to the MCP Bridge repository (git)

Installation steps:

1. Clone the repository: git clone https://github.com/mahawi1992/mwilliams_mcpbridge.git cd mwilliams_mcpbridge

2. Install dependencies: npm install

3. Copy example config to a working config file and edit it: cp mcpbridge.config.example.json mcpbridge.config.json

   Edit mcpbridge.config.json to point to your MCP servers and set any needed environment variables

4. Start the bridge server (example): node bridge-server.js

5. Verify setup with the provided tooling or your client configuration.

Optional configuration (examples):

• Claude integration via CLI or desktop by pointing to the bridge server as shown in the setup section of the README.
• Use of mcpbridge.config.json to centrally manage server definitions and environment variables.

Tips and common notes:

• The bridge uses lazy schema loading, so tool schemas are fetched only when requested to minimize initial context usage.
• Results larger than 2KB are automatically compacted and stored externally; use get_result with the returned reference ID to retrieve full data.
• Ensure absolute paths for bridge-server.js to avoid runtime path resolution issues.
• If tools or servers fail to load, check connectivity to underlying MCP backends and verify that npm install completed successfully.
• Restart clients (Claude Desktop/Code) after updating the bridge configuration to ensure new servers are discovered.