smithsonian

This MCP server exposes the Smithsonian Open Access API through a local MCP endpoint that Claude can query. It enables searching the Smithsonian collections, retrieving detailed item metadata, and exploring available category terms to refine searches. The available tools include search_collection for flexible queries, get_item for item-level details, and get_category_terms to discover facets like online_media_type, data_source, topic, and place. Configure the server with your Smithsonian API key, then connect Claude Desktop to the smithsonian MCP server so Claude can issue these tools on your behalf. The server handles API authentication and translates user requests into API calls, returning structured results that Claude can present.

Prerequisites:

• Python 3.10 or higher
• A Smithsonian API key (free) obtained from https://api.data.gov/signup/

Step-by-step installation:

1. Clone the repository and install:

git clone https://github.com/AlexLin1234/smithsonian-mcp-server.git
cd smithsonian-mcp-server

# Install in editable mode (preferred for development)
pip install -e .

2. Configure the API key:

cp .env.example .env
# Edit .env and add your API key if required by the project

Alternatively, set the environment variable directly:

export SMITHSONIAN_API_KEY=your_api_key_here

3. Run the MCP server locally to verify:

python -m smithsonian_mcp_server.server

4. Connect Claude Desktop using the provided configuration example (see next section) to start issuing MCP commands.

Notes and tips:

• Ensure your SMITHSONIAN_API_KEY is available in the environment where the MCP server runs.
• If you’re behind a firewall or proxy, make sure outbound API requests toapi.data.gov are allowed.
• The server relies on the Smithsonian Open Access API; if the API changes or the key is invalid, requests will fail with authentication or rate-limit messages.
• When integrating with Claude Desktop, you can place the smithsonian entry in Claude’s mcpServers config with the Python invocation shown in README; you can also point to a virtual environment’s Python executable if you prefer isolation.
• For development, run tests or linting as documented in the repository to catch formatting or compatibility issues early.