Use AstroFetch from your own AI agent

If you already work with an AI coding assistant such as Claude Code, Claude Desktop, OpenCode or Codex, you can connect it to AstroFetch. Your assistant then gets the same curated archive knowledge the AstroFetch chat uses (table guides, column documentation and worked example queries), and can run ADQL queries against NASA archives, look up literature in NASA ADS, and plan observations, all from inside your own tools. It uses the open Model Context Protocol (MCP), so no AstroFetch account or API key is needed.

The address

Every client below needs just this one URL:

https://astrofetch.ipac.caltech.edu/mcp

There is nothing to install and nothing to sign in to. Opening that address in a web browser brings you back to this page; it is meant for your assistant, not for a browser.

Set up your client

Claude Code

Run this once in a terminal:

claude mcp add --transport http --scope user astrofetch https://astrofetch.ipac.caltech.edu/mcp

--scope user makes it available in all your projects. Leave it out to add it to the current project only. To share it with collaborators through a project's repository, put this in the project's .mcp.json instead:

{
  "mcpServers": {
    "astrofetch": {
      "type": "http",
      "url": "https://astrofetch.ipac.caltech.edu/mcp"
    }
  }
}

Claude Desktop and claude.ai

Add it as a custom connector. Go to Customize, then Connectors, click +, then Add custom connector. Paste the address above and click Add. Leave the OAuth fields under Advanced settings empty. The Free plan allows one custom connector. On Team and Enterprise plans an organization owner adds custom connectors instead, under Organization settings, then Connectors.

OpenCode

Add this to your opencode.json:

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "astrofetch": {
      "type": "remote",
      "url": "https://astrofetch.ipac.caltech.edu/mcp",
      "enabled": true
    }
  }
}

Codex

Run this once:

codex mcp add astrofetch --url https://astrofetch.ipac.caltech.edu/mcp

or add these lines to ~/.codex/config.toml:

[mcp_servers.astrofetch]
url = "https://astrofetch.ipac.caltech.edu/mcp"

Other clients

Any client that can connect to a remote MCP server over Streamable HTTP will work. Give it the address above; no authentication is required.

What you can ask for

Ask in plain language, as you would a colleague. For example:

• “Using AstroFetch, find all confirmed transiting planets with radius under 2 Earth radii and orbital period under 10 days, and save them to a CSV.”
• “Which Keck instruments observed M31 in 2024? Use the Keck Observatory Archive through AstroFetch.”
• “When are the next three transits of HD 189733 b visible from Palomar?”
• “Find recent papers on the TRAPPIST-1 atmospheres and summarize what JWST has found.”

Through this connection your assistant can reach the NASA Exoplanet Archive, IRSA (the NASA/IPAC Infrared Science Archive) and the Keck Observatory Archive. It can also search NASA ADS, predict exoplanet transits, and compute when a target is up from a named observatory.

As in the chat, treat results as drafts: ask your assistant to show you the ADQL it ran, and check it before relying on the answer.

How results come back

A small result (up to about 1,000 rows) comes back to your assistant directly, as a table it can read and save.

A larger result comes back as its size, a preview of about 20 rows so your assistant can check the columns, and a download link. The download comes straight from the archive, not through AstroFetch. The link stays valid for about 24 hours; download what you need soon after the query finishes.

Some queries take a while at the archive. When one does, your assistant is told the query is still running and checks back until it finishes. That is normal, not an error.

Limits and fair use

The service is free and anonymous. To keep it available for everyone, each IP address has a rate limit on queries and lookups; ordinary interactive use will not come near it. Literature search runs on a shared NASA ADS key with a daily limit; if that is used up for the day, your assistant will say so, and you can get your own free key from NASA ADS.

Please do not use it for bulk scraping. For very large pulls, ask your assistant for the ADQL and run it yourself against the archive's own TAP service.

If results from AstroFetch contributed to a paper, please acknowledge AstroFetch and cite each archive whose data you used.

Questions

Email astrofetch-help@ipac.caltech.edu. We would especially like to hear which clients you use and where the connection falls short.