2025 workshop program
AstroAgent: An Intelligent Assistant for Astronomical Research Based on MCP Protocol
The work
Abstract
Astronomical research workflows often begin with time-consuming literature surveys and manual data curation processes. While artificial intelligence-powered agent systems have emerged as potential solutions, existing implementations predominantly rely on astronomy-specific large language models (LLMs) that demand substantial annotated datasets and computationally expensive fine-tuning procedures.
In this study, we develop AstroAgent, an innovative intelligent assistant system for astronomical research that leverages the Model Context Protocol (MCP) framework and state-of-the-art reasoning models. Our system needs no specialized astronomical LLMs while providing extendibility to more astronomical tools. The current implementation incorporates three dedicated MCP components: (1) an ADS MCP Server for natural language-based literature retrieval and analysis, (2) a SIMBAD MCP Server for intelligent access to stellar catalog data referenced in publications, and (3) a Python Documentation MCP Server that automatically generates customized data processing scripts through natural language instructions.
This system enables researchers to conduct comprehensive literature reviews, monitor field developments, perform cross-dataset data synthesis, and generate analysis scripts through intuitive conversational interfaces. By automating these routine tasks, AstroAgent allows astronomers to focus on higher-level scientific interpretation while maintaining direct access and customized process to distributed astronomical data. The system’s modular architecture supports both customization for specific research needs and future integration of additional astronomical tools through standardized MCP interfaces.