papersgpt-for-zotero

PapersGPT for Zotero exposes an MCP server that bridges your Zotero library with a range of AI models and prompts through the PapersGPT plugin ecosystem. The MCP server enables chat and analysis capabilities directly from your Zotero content, supporting fast full-text search, summarization, extraction of key insights, and structured notes insertion back into Zotero. When you connect via MCP, you can use compatible chatbot clients to query Zotero items, run built-in prompts (e.g., summaries, literature reviews, theoretical frameworks), and leverage local or remote LLMs configured in your client to process PDFs and metadata. The integration is designed to be lightweight and fast, emphasizing seamless interaction with Zotero notes and collections while minimizing data transfer to remote services.

To use it, start the MCP server (for example via Docker as described in the installation steps) and point your MCP-enabled client to the server host. In Zotero, enable the PapersGPT plugin and ensure the SSE endpoint at http://localhost:9080/sse is reachable by the MCP host. Then, in your chatbot client, connect to the MCP server using the server name papersgpt-for-zotero and begin issuing queries such as “summarize key findings from [paper], focusing on methodology” or “extract data points on sample size and outcomes.”

Prerequisites:\n- Docker installed on your system (Windows, macOS, or Linux)\n- Zotero installed with the PapersGPT plugin (as described in the PapersGPT quickstart)\n- Access to the internet for pulling the MCP Docker image (if using the docker approach)\n\nStep-by-step:\n1) Install Docker: follow the official instructions for your OS (https://docs.docker.com/get-docker/).\n2) Pull or build the PapersGPT for Zotero MCP image:\n - If an official image is published: docker pull papersgpt/papersgpt-for-zotero\n - If you have a local image: docker build -t papersgpt-for-zotero .\n3) Run the MCP server:\n docker run -i papersgpt/papersgpt-for-zotero\n Note: If your environment requires ports, consider exposing or mapping the SSE endpoint (for example, -p 9080:9080) and ensure the Zotero plugin can reach http://localhost:9080/sse as described in the server docs.\n4) Configure the Zotero/PapersGPT plugin to communicate with the MCP server, and ensure the SSE URL is reachable by your client.\n5) Verify connectivity by sending a basic query from your MCP-enabled client (e.g., check server status or perform a simple search).\n\nIf you prefer a non-Docker installation, follow the vendor-specific instructions for building or running the native C++ MCP server binary and configure the same environment variables as described in the mcp_config section.

Tips and known considerations:\n- The MCP server is designed to be lightweight and fast, focusing on Zotero integration and local processing. Make sure your Zotero notes and attachment data are within the scope of the plugin’s permissions.\n- Ensure the SSE endpoint (http://localhost:9080/sse) is accessible from your MCP host and firewall rules allow local traffic on the port.\n- If you encounter authentication prompts from Zotero or model providers, configure API keys or OAuth tokens in the PapersGPT plugin as directed in the Quickstart guides.\n- When using Docker, you may need to adjust memory and CPU allocations to optimize performance for large document sets.\n- The server supports a range of LLMs via the client; choose models appropriate to your data privacy and cost constraints.\n- For troubleshooting, verify that the environment variables in mcp_config (ZOTERO_HOST, ZOTERO_PLUGIN_PATH, MCP_SSE_URL) are correctly set to reflect your local setup.