openalex

This MCP server provides a set of tools to interact with the OpenAlex scholarly works database via the MCP protocol. The server exposes five tools: search_works, get_work_details, get_referenced_works, get_citing_works, and get_work_ngrams. You can use these tools through an MCP client to query for papers by keywords, fetch detailed metadata for a specific work, retrieve the references a work cites, find works that cite a given paper, and obtain N-gram information for a work's full text when available. Keep in mind that OpenAlex generally does not store full text due to copyright, so full-text search is limited to title and abstract per OpenAlex capabilities. To use the tools, configure the openalex MCP server in your client (e.g., Claude integration) and call the desired tool with the required parameters as shown in the README examples.

Prerequisites: Python and the uvx CLI installed, and access to a machine where you can run the server.

1. Clone the repository (optional): git clone <repository-url> cd openalex-mcp-server

2. Create and activate a Python virtual environment, then install dependencies: uv venv source .venv/bin/activate

   uv pip install . This installs the server package along with its dependencies (mcp[cli], pyalex).

3. Run the server locally (development): uv --directory /YOUR/INSTALL/DIRECTORY/openalex-mcp-server run server.py

4. Set up the OpenAlex email for polite pool usage (optional but recommended): export OPENALEX_EMAIL="your.email@example.com"

Tips and considerations:

• To maximize throughput, set OPENALEX_EMAIL to enable the faster OpenAlex polite pool; without it, the server will use the anonymous pool with stricter rate limits.
• When configuring MCP client integrations, ensure the environment variable OPENALEX_EMAIL is present in the client configuration if you rely on the polite pool.
• The get_work_ngrams endpoint provides N-gram data only if available; many works will not have N-gram data.
• If you experience timeouts or rate limiting, consider adjusting the client timeout (e.g., 60s) and respecting OpenAlex rate limits by using filters and pagination thoughtfully.