home / mcp / sdk mcp server

SDK MCP Server

Provides searchable access to multiple SDK sources with tools to list, view, search, and compare OpenAI Agents SDK and Google ADK implementations.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "arunmenon-sdk-mcp-server": {
      "command": "python",
      "args": [
        "/absolute/path/to/sdk-mcp-server/src/server.py"
      ],
      "env": {
        "PYTHON": "/usr/bin/python3"
      }
    }
  }
}

You can run an MCP server that gives you fast, searchable access to multiple SDK sources (such as OpenAI Agents SDK and Google ADK) so you can compare implementations, inspect source code, and find the right method signatures across SDKs. This server is designed to be easy to extend and integrates with clients that support MCP endpoints.

How to use

To use the SDK MCP Server with your MCP client, first ensure the server is running locally or on a reachable host. The client connects to the server to issue tool-specific queries like listing files, retrieving source code, searching code, or extracting full class definitions across configured SDKs. You can request cross-SDK comparisons and example patterns to speed up debugging and learning across OpenAI Agents SDK and Google ADK.

How to install

Prerequisites you need before installing: - Python 3.8 or later - internet access to download dependencies - access to a shell or terminal Follow these concrete steps to install and run the MCP server.

# 1) Install Python dependencies
pip install -r requirements.txt

# 2) Download SDK sources (downloads configured SDKs as defined in sdks.yaml)
python src/download_sdk.py

# 3) Start the MCP server locally (example path; adjust to your environment)
# You will configure the client to connect to this local server
python /absolute/path/to/sdk-mcp-server/src/server.py

Configuration and startup notes

Configure your MCP client to connect to the server using the provided stdio command. The typical setup shown for Claude Desktop uses a Python runtime to launch the server script from your local path.

Additional notes

The server exposes a set of tools grouped by SDK. You can list files, fetch source code, search within code, extract full class definitions, and find usage examples for each configured SDK. The server is designed to make it easy to extend with additional SDKs by updating the configuration.

Available tools

openai_agents_list_files

List all available OpenAI Agents SDK source files.

openai_agents_get_source

Get source code of a specific OpenAI Agents SDK file.

openai_agents_search_code

Search for terms, methods, or patterns within the OpenAI Agents SDK sources.

openai_agents_get_class

Extract the complete definition of a specified class from the OpenAI Agents SDK.

openai_agents_find_examples

Find usage examples and patterns within the OpenAI Agents SDK.

google_adk_list_files

List all available Google ADK source files.

google_adk_get_source

Get source code of a specific Google ADK file.

google_adk_search_code

Search for terms, methods, or patterns within the Google ADK sources.

google_adk_get_class

Extract the complete definition of a specified class from the Google ADK.

google_adk_find_examples

Find usage examples and patterns within the Google ADK.

list_available_sdks

Show all configured SDKs across the MCP setup.

search_all_sdks

Search across all configured SDKs for a given query.

compare_implementations

Compare similar concepts across multiple SDKs to highlight differences.