Auto MCP

Visit the Auto MCP Homepage

Transform any OpenAPI/Swagger definition into a fully-featured Model Context Protocol (MCP) server – ready to run locally, inside Claude Desktop, or in the cloud.

The service reads a Swagger (OpenAPI v2) or OpenAPI v3 document, generates routes on-the-fly, proxies requests to the upstream endpoint you configure, and exposes them through MCP using either the STDIO or HTTP or SSE transport defined in the MCP specification.

✨ Why Auto MCP?

• Zero boiler-plate – bring your `swagger.json` and start serving.
• Flexible deployment – run as a CLI, long-lived daemon, or within Docker/Kubernetes.
• All transport modes –
  • `stdio` (default).
  • `http` - StreamableHttp, newest MCP prototcol.
  • `sse` – self-hosted long-running event source.
• Pluggable auth – bearer token, basic auth, API keys, OAuth2 or no auth.
• Runtime configuration – YAML file, CLI flags, or environment variables (prefixed `AUTO_MCP_`).

🛠️ Using Auto MCP

Easily tailor your Swagger/OpenAPI file for optimal MCP integration. The MCP Config Builder lets you:

• Edit endpoint descriptions for clearer, more helpful documentation.
• Filter out unnecessary routes to streamline your API exposure.
• Preview and customize how endpoints appear to LLMs and clients.
• Generate an adjustment file (`--adjustment-file`) for use with Auto MCP, applying your customizations automatically.

How it works

1. Install the MCP Config Builder:

go install ./cmd/mcp-config-builder

This will build and install the `mcp-config-builder` binary to your `$GOPATH/bin` (usually `~/go/bin`). Make sure this directory is in your `PATH`.

2. Launch the tool:

mcp-config-builder --swagger-file=/path/to/swagger.json

3. Interactively review and edit endpoints in a user-friendly TUI (Terminal User Interface).

4. Save your adjustments to a file for future use or sharing.

5. Run Auto MCP with your adjustment file to apply your customizations:

auto-mcp --swagger-file=/path/to/swagger.json --adjustment-file=/path/to/adjustments.json

📚 Use Cases

1. Rapid Prototyping: Wrap any REST API as an MCP server in seconds—ideal for testing ideas or building AI tools fast.

2. Bridge Legacy Services: Expose legacy or internal systems as MCP endpoints without rewriting them.

3. Access Any 3rd-Party API from Chat Applications: Turn any third-party API into an MCP tool, making it accessible to AI assistants like Claude.

4. Minimal Proxy Tools: Use auto-mcp to proxy APIs that already handle validation and logic—no wrappers needed.

🖥️ Running inside Claude Desktop

Add the following snippet to your Claude Desktop configuration (⟂ _Settings → MCP Servers_):

{
  "mcpServers": {
    "YourMCP": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "-v",
        "/Users/you/path/to/swagger.json:/server/swagger.json",
        "ghcr.io/brizzai/auto-mcp:latest"
        "--swagger-file=/server/swagger.json"
      ]
    }
  },
  "globalShortcut": ""
}

Claude will start the container on-demand and connect over STDIO. Replace the host path to `swagger.json` and image tag to suit your setup.

CLI flags

• `--mode` – override `server.mode` (`stdio` or `sse`).
• `--swagger-file` – path to the OpenAPI document (default: `swagger.json`).
• `--adjustment-file` - mcp-config-builder output filter/change route descriptions

For detailed configureation guidelines, please see CONFIGURATION.md.

🔐 OAuth Support

Auto MCP supports OAuth 2.1 authentication, including PKCE, dynamic client registration, and multiple providers (internal, GitHub, Google). This allows you to secure your MCP server with industry-standard authentication flows.

See the OAuth Usage Guide for detailed setup instructions, endpoint descriptions, and testing tips.

🐳 Running with Docker

1. Run in local `stdio` mode:

docker run --rm -i \
     -v $(pwd)/swagger.json:/server/swagger.json \
       ghcr.io/brizzai/auto-mcp:latest \
       --swagger-file=/server/swagger.json \
       --mode=stdio

2. Run in remote `sse`/`http` mode :

docker run \
     -v $(pwd)/swagger.json:/server/swagger.json \
       ghcr.io/brizzai/auto-mcp:latest \
       --swagger-file=/server/swagger.json \
       --mode=http

The bundled `docker-compose.yml` maps port 8080 and persists logs to `./logs`.

Running the Petshop Example

You can try out the included Petshop demo using Docker. This demo uses a sample configuration and API specs to show how auto-mcp works.

Steps:

1. Make sure you are in the root directory of this repository.

2. Run the following command:

docker run --rm -i \
  -v $(pwd)/examples/petshop/config:/config \
  ghcr.io/brizzai/auto-mcp:latest

• This command mounts the `examples/petshop/config` directory from your local machine into the Docker container at `/config`.
• The `/config` directory inside the container should contain:
  • `config.yaml`: Main configuration file for the demo
  • `swagger.json`: API specification for the Petshop service
  • `adjustment.yaml`: Optional adjustments or overrides for the API

> Note:

> Any files you place in `examples/petshop/config` will overwrite the default `config.yaml`, `swagger.json`, and `adjustment.yaml` inside the container.

This setup allows you to easily test and modify the Petshop demo configuration.

See [docs/CONFIGURATION.md](docs/

CONFIGURATION.md) for all config options

and environment variable overrides.

🤝 Contributing

For detailed contribution guidelines, please see CONTRIBUTING.md.

📄 License

Distributed under the Apache License 2.0 License. See `LICENSE` for details.