Search astronomical catalogs (SIMBAD, JPL Horizons, NED, VizieR) via astronomy-mcp (uvx)

astronomy-mcp (uvx) — query SIMBAD, JPL Horizons, NED, VizieR, TNS from an agent

Package: astronomy-mcp v0.1.4 on PyPI Transport: stdio via uvx astronomy-mcp Auth: none — all backends (SIMBAD, JPL Horizons, NED, VizieR, TNS) are free public APIs Tools: 9

Tool inventory

Recipe

1.  uvx astronomy-mcp                                    # start server
2.  → initialize                                          # MCP handshake
3.  → notifications/initialized
4.  → tools/call  search_simbad_object                    # look up a deep-sky object
       { "query": "Andromeda Galaxy" }

Real trace (Andromeda Galaxy via SIMBAD)

Request:

{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"search_simbad_object","arguments":{"query":"Andromeda Galaxy"}}}

Response:

Found 1 matches:
- ID: M  31
  Name: M  31
  Type: AGN
  RA: 10.68471, Dec: 41.26875
  Magnitude: 3.44

Composable patterns

1. "Is Mars visible from Istanbul tonight?" — chain geocode-mcp (get Istanbul coords: 41.006, 28.976) → search_horizons_object("Mars") → check_visibility(object_name="Mars", latitude=41.006, longitude=28.976, observation_time="now").

1. "What's near the Crab Nebula?" — search_simbad_object("Crab Nebula") to get RA/Dec, then cone_search(ra=83.63, dec=22.01, radius=0.5, catalog="simbad").

1. "Find dark matter catalogs" — search_vizier_catalogs("dark matter") to discover relevant survey datasets.

Failure modes

• Queries hit real public APIs — expect network latency (1–5s per call).
• SIMBAD classifies M31 as "AGN" (active galactic nucleus) which is technically correct but may surprise users expecting "galaxy". The type taxonomy is SIMBAD's, not the server's.
• search_horizons_object for comets: use provisional designation format "YYYY XX" (e.g., "2025 N1"), NOT "C/2025 N1 ATLAS". The full designation often fails.

When to use

Science education agents, astrophotography planning, observatory scheduling, research literature grounding (resolve object names to coordinates before querying papers), or any agent that needs to talk about the sky with real data behind it.