Bridge MCP Server
OpenAI API interfaces bridged to Model Context Protocol tools for broader compatibility.
Data last scanned today · Reviewed today
Overview
MCP-Bridge functions as an adapter, enabling standard OpenAI API clients to interact with MCP tools. By providing compatible endpoints, it allows users to leverage MCP functionalities in environments that do not natively support the protocol. It handles chat completions and MCP tool calling through a Dockerized service. The project supports both streaming and non-streaming completions. It is designed to work with inference engines that support tool calling, such as vLLM or Ollama, and requires a JSON-based configuration to define the available MCP servers.
Our verdict
With a security score of 39/100, this project is classified as high-risk. It is community-maintained rather than vendor-backed, and its last commit was over 240 days ago. Because the authentication methods and read-only support have not been reviewed, and the maintainer has marked the project as soft deprecated in favor of native Open WebUI support, this tool should only be used in isolated, experimental environments.
- Supports both streaming and non-streaming chat completions.
- Zero direct dependencies reduces the supply-chain surface area.
- Permissive MIT license allows for open modification.
- Integrated support for MCP sampling and tools.
- High security risk due to lack of editorial reviews.
- Project is soft deprecated in favor of native solutions.
- Stale codebase with no commits in over 240 days.
- Authentication and read-only modes remain unverified.
Compatibility
| Client | Local | Docker | Remote | Read-only |
|---|---|---|---|---|
| Claude Desktop | ||||
| Cursor | ||||
| VS Code | ||||
| Windsurf | ||||
| ChatGPT |
Frequently asked questions
›Is MCP-Bridge still being actively developed?
MCP-Bridge is considered soft deprecated because Open WebUI now natively supports the Model Context Protocol as of version 0.6.31. The project is currently looking for new maintainers to assist with future development.
›What kind of inference engine do I need to run this?
You must use an inference engine that supports tool calls, such as vLLM or Ollama.
›How do I configure additional MCP servers?
You add new servers by editing the config.json file and defining the command and arguments required to execute them.
›Does MCP-Bridge support authentication?
Yes, you can enable API key authentication by adding an auth section with api_keys to your config.json file. If no API key is configured, authentication is skipped.
›Can I use this with clients that require STDIO for MCP?
MCP-Bridge uses SSE for communication, so for STDIO-only clients like Claude Desktop, you would need to use an additional tool such as mcp-gateway.
›How can I verify that my MCP server configuration is working?
You can visit the documentation at /docs on your server to list available tools, or use an SSE client like mcp-cli to connect to the SSE endpoint at /mcp-server/sse.
Badge
Maintain this server? Add the live badge to your README.