mcp-proto-okn: Natural-language access to open scientific knowledge graphs through the Model Context Protocol

The paper introduces mcp-proto-okn, a tool designed to bridge the gap between complex scientific knowledge graphs and AI assistants. By leveraging the Model Context Protocol (MCP), this system allows researchers to interact with vast, cross-domain biomedical and scientific data using natural language. The primary goal is to lower the technical barriers that often prevent scientists from performing sophisticated cross-domain analysis on large-scale knowledge graphs.

How it works

The system is built as a Python-based MCP server utilizing the FastMCP framework. It acts as an intermediary that translates natural language requests from an AI assistant into actionable operations on scientific knowledge graphs. By integrating directly with the Model Context Protocol, it enables AI models to perform a variety of complex data tasks, such as routing queries to the appropriate graph, inspecting schemas, and executing SPARQL queries.

Key capabilities

The server provides a suite of features that facilitate deep scientific inquiry, including:

• Multi-graph querying: The ability to pull information from multiple scientific knowledge sources simultaneously.

• Ontology expansion: Tools to help users broaden their search parameters and understand the relationships within the data.

• Transcript generation: A feature that documents the analysis process, ensuring that the steps taken by the AI are transparent and reproducible.

• Schema inspection: Allows the AI to understand the structure of the underlying data, which is essential for accurate query formulation.

Accessibility for researchers

A significant focus of this project is to make scientific data analysis more accessible to users who may not be experts in database query languages like SPARQL. By enabling natural-language access, the authors aim to streamline the workflow for biomedical and scientific researchers. The project is open-source, with the authors providing full documentation, configuration instructions for clients, and example transcripts in their GitHub repository to help users get started with their own analyses.