This example demonstrates how to run a Model Control Protocol (MCP) server that integrates with Codegen. The server provides:

1. A standardized interface for model inference
2. Integration with Codegen's core functionality, parsing codebases and executing codemods
3. Support for various LLM providers through the MCP protocol

• uv
• python 3.12+

1. No installation is necessary, with the following command. Run this command directly or add it your .json mcp config file.

 uvx --from 'git+https://github.com/codegen-sh/codegen-sdk.git#egg=codegen-mcp-server&subdirectory=codegen-examples/examples/codegen-mcp-server' codegen-mcp-server

Here is an example mcp config that can be used with Cline or Claude desktop to integrate this MCP server

{
  "mcpServers": {
    "codegen-mcp": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/codegen-sh/codegen-sdk.git#egg=codegen-mcp-server&subdirectory=codegen-examples/examples/codegen-mcp-server",
        "codegen-mcp-server"
      ],
      "disabled": false,
      "autoApprove": []
    }
  }
}

• parse_codebase: Parses a codebase located at the provided path.
• check_parse_status: Provides the current parsing status for the provided codebase.
• execute_codemod: Executes a codemod script on a parsed codebase. This is where the codegen sdk leveraged to run simple or sophisticated codemods on the codebase.