astro-orchestra

Astro Orchestra - Multi-Agent Astronomy Research System

A sophisticated multi-agent system for conducting astronomy research using the Model Context Protocol (MCP) to orchestrate specialized agents and external astronomical data tools.

Quick Start

1. Clone and Setup:

cd astro-orchestra
python -m venv venv
source venv/bin/activate  # or `venv\Scripts\activate` on Windows
pip install -r requirements.txt

1. Configure Environment:

# Create .env file with your Google API key
echo "GOOGLE_API_KEY=your_google_api_key_here" > .env

1. Run MCP Server:

python -m src.mcp.server

1. Configure Cursor:

Add to your .cursor/mcp_config.json: ``json { "mcpServers": { "astronomy-research": { "command": "python", "args": ["-m", "src.mcp.server"], "cwd": "/path/to/astro-orchestra", "env": { "PYTHONPATH": ".", "GOOGLE_API_KEY": "${env:GOOGLE_API_KEY}" } } } } ``

Project Architecture

%%{init: {'flowchart': {'nodeSpacing': 50, 'rankSpacing': 80}}}%%
graph TD
    %% Styling with darker text
    classDef human fill:#FFE4B5,stroke:#FF8C00,stroke-width:3px,color:#333
    classDef orchestrator fill:#E6E6FA,stroke:#9370DB,stroke-width:3px,color:#333
    classDef specialist fill:#B0E0E6,stroke:#4682B4,stroke-width:2px,color:#333
    classDef tool fill:#90EE90,stroke:#228B22,stroke-width:2px,color:#333
    classDef langgraph stroke:#4169E1,stroke-width:3px
    classDef mcp stroke:#32CD32,stroke-width:3px,stroke-dasharray: 5 5
    classDef feedback stroke:#FF6347,stroke-width:3px

    %% Top level - Human Interface
    HC[Human + Cursor]:::human
    
    %% Core Orchestrator
    HC -->|"Research Query"| O[Orchestrator Agent]:::orchestrator
    
    %% Specialist Agents arranged in a row
    O -->|"Routes Tasks"| P[Planning Agent]:::specialist
    O --> DG[Data Gathering Agent]:::specialist
    O --> A[Analysis Agent]:::specialist
    O --> TS[Theorist Simulation Agent]:::specialist
    O --> LR[Literature Reviewer Agent]:::specialist
    
    %% Return paths
    P -->|"Return Results"| O:::langgraph
    DG --> O:::langgraph
    A --> O:::langgraph
    TS --> O:::langgraph
    LR --> O:::langgraph
    
    %% MCP Tool Servers - bottom level
    DG -.->|"MCP Protocol"| DESI[DESI Server]:::tool
    DG -.-> LSST[LSST Server]:::tool
    DG -.-> CMB[CMB Server]:::tool
    A -.-> STATS[Statistics Server]:::tool
    TS -.-> NBODY[N-body Server]:::tool
    LR -.-> ARXIV[ArXiv Server]:::tool
    
    %% Human Feedback Loop
    O ==>|"Pauses for Review<br/>Shows Results"| HC:::feedback
    HC ==>|"Provides Feedback"| O:::feedback

Astro Orchestra uses a multi-agent architecture where:

• Orchestrator Agent: Breaks down research tasks and coordinates specialists
• Data Gathering Agent: Accesses DESI, LSST, CMB databases via MCP tools
• Analysis Agent: Performs statistical analysis and computations
• Theorist Simulation Agent: Runs cosmological simulations
• Literature Reviewer Agent: Searches and synthesizes scientific papers

All agents communicate through a shared state system and use external MCP tool servers for accessing astronomical data and computational resources.

MCP Integration

The system exposes itself as an MCP server that can be integrated with AI assistants like Cursor. External astronomy tools run as separate MCP servers, allowing for:

• Dynamic tool discovery
• Scalable tool ecosystem
• Separation of concerns
• Easy tool updates and additions

Development Status

This project is in active development. Core components are being implemented incrementally.

License

MIT License