Mcpo MCP Server

Expose MCP tools as OpenAPI-compatible HTTP servers for RESTful agent integration.

Data last scanned today · Reviewed today

63★ 4,315PythonMIT

Overview

Mcpo functions as a proxy that transforms raw MCP server commands into standard HTTP-based OpenAPI endpoints. By removing the requirement for raw stdio or custom socket handling, it allows AI agents and applications to interact with MCP tools using familiar web standards. It includes features for auto-generating interactive documentation for every connected tool, ensuring that AI clients can discover and use tool functions without manual configuration. Developers can deploy this server using command-line arguments to specify the MCP command or endpoint, supporting both local stdio-based servers and remote SSE or streamable HTTP instances. It offers advanced configurations such as serving under custom subpaths for reverse proxy compatibility and managing multiple tools through a single configuration file.

Our verdict

With a security score of 53/100, this project is classified as high-risk. The score is impacted by the fact that core mechanisms like authentication methods and read-only mode support remain unreviewed. As a community-maintained project that is not officially reviewed, it should be treated as experimental. Users should exercise caution in production environments where secure handling of tool outputs is critical, as the current implementation's security posture is not yet established by audit.

Zero direct dependencies minimize the supply-chain attack surface.
Auto-generates interactive API documentation for connected tools.
Supports multiple server types including SSE and streamable HTTP.
MIT licensed with 4315 GitHub stars indicating community adoption.

Authentication method security is not yet formally reviewed.
Lack of editorial review on the read-only mode implementation.
Community-maintained without official vendor oversight or audits.
Project has not received an independent security review.

Compatibility

Client	Local	Docker	Remote	Read-only
ChatGPT
Claude Desktop
Cursor
VS Code
Windsurf

Frequently asked questions

›Can I run multiple MCP tools through a single mcpo instance?

Yes, you can use a configuration file following the Claude Desktop format to serve multiple MCP tools simultaneously under unique routes.

›Does mcpo support MCP servers that use non-stdio communication?

Yes, mcpo supports stdio, SSE, and streamable HTTP server types by specifying the --server-type flag or defining it in a configuration file.

›Is it possible to update my MCP configuration without restarting the server?

Yes, if you run mcpo with the --hot-reload flag, it will automatically watch your configuration file for changes and reload the servers without downtime.

›How can I secure access to my proxied MCP tools?

You can protect your instance by providing an --api-key argument when starting the server.

›Does mcpo provide documentation for the tools it exposes?

Yes, mcpo automatically generates interactive documentation for every tool, which can be accessed via the /docs endpoint.

›Can I host mcpo behind a reverse proxy or on a specific subpath?

Yes, you can use the --root-path argument to serve all routes under a specific path, such as /api/mcpo.