SkyQuery
A local, open-source MCP server that wraps astronomy data services (SIMBAD, JPL Horizons, VizieR, etc.) behind a conversational interface, enabling AI assistants to answer astronomy questions with real, typed, and provenance-tracked data.
README
<div align="center">
✦ SkyQuery
The sky, queryable. Your questions, real data, one conversation.
A free, local, open-source MCP server that puts the working astronomer's whole toolkit, object lookups, catalog cross-matches, ephemerides, literature, and observation planning, behind one conversational interface. Point any AI assistant at it and ask the sky anything. Runs entirely on your machine.
Website · Install · The demo · How it works · Tools
</div>
The honest version, first
SkyQuery is a normalization and MCP layer on top of astroquery, which it credits loudly. It wraps free, publicly funded astronomy services (SIMBAD, JPL Horizons, VizieR, Gaia, ADS, and more), honors their rate limits, and attaches a citation to every value it returns. It is local-first: there is no SkyQuery server, nothing phones home, and your queries, keys, and downloaded data never leave your machine except to talk to those public services directly. You can read every line that touches the network.
It is not affiliated with NASA, JPL, CDS/Strasbourg, STScI, ESA, the Astropy project, or the unrelated JHU catalog tool of a similar name. It is not proprietary-data access, not telescope control, and not a hosted service. The answer to "can you host it for me" is "run it yourself, that is the point."
Why it exists
The data is public and abundant. The interfaces are fragmented and unforgiving. To answer one ordinary question, "where is comet Apophis tonight, how big is it, and what's the latest paper on it?", you touch a JPL prompt, a SIMBAD form, and ADS query syntax, each with its own units and conventions, and you end up copying numbers between tabs by hand. Ask an LLM directly and it will happily invent an ephemeris that looks right and is wrong.
SkyQuery is the missing layer. It hands the assistant results that are typed, unit-tagged, and provenance-carrying instead of raw floats, so the model can reason over real data without guessing what "deg versus hourangle" or "J2000 versus ICRS" means.
The demo that sells it
"Where is asteroid Apophis on its 2029 approach, how big is it, and what's a paper about it?"
$ skyquery demo
99942 Apophis (2004 MN4)
size: 0.34 km H: 19.09 mag
closest approach: 2029-Apr-13 22:00 0.000257245 AU (0.100 lunar distances)
visual magnitude at approach: 4.257 mag
paper: Radar and Optical Observations of Near-Earth Asteroid 99942 Apophis (2024)
99942 Apophis (2004 MN4) is about 0.34 km. On 2029-Apr-13 22:00 it passes 0.000257 AU
from Earth (0.10 lunar distances), inside geostationary orbit. Source: JPL Horizons and SBDB.
JPL Horizons, the Small-Body Database, and the literature, answered in one breath, with the units, the frame, and the citation intact. That runs offline against shipped fixtures, so it works the moment you install it.
Install in under five minutes
# 1. Install (pick one)
uv tool install skyquery-mcp # recommended
pipx install skyquery-mcp
# 2. Try it right now, offline, no keys
skyquery demo
skyquery resolve Vega
skyquery ephemeris "99942 Apophis" --start 2029-04-13 --stop 2029-04-14
Then point your assistant at it. Add one stdio block to your MCP client config (for example Claude
Desktop's claude_desktop_config.json):
{
"mcpServers": {
"skyquery": { "command": "skyquery-mcp" }
}
}
That is the whole integration. SIMBAD, VizieR, and JPL Horizons need no key at all, so it is useful out of the box. Optional free keys unlock more, and go straight to your OS keychain, never a file:
skyquery login ads # free ADS researcher token, unlocks literature
skyquery login nasa # free NASA key, unlocks the APOD "wonder" layer
See SETUP.md for the exact steps and the --live flag.
How it works
Your assistant ──MCP──▶ SkyQuery ──▶ SIMBAD · Horizons · VizieR · Gaia · ADS · ...
◀── normalized + provenance ──
Every service response is parsed into one small set of typed models, Object, CatalogTable,
Ephemeris, Paper, ObservationWindow, DataProduct, with values wrapped in astropy units and
coordinates carried in explicit frames and epochs. One source of truth feeds both the MCP tools and
the CLI.
- One coordinate frame. Everything lands in ICRS degrees with its epoch explicit.
- Every value sourced. Each field knows the service, the exact query, and the acknowledgment
that service asks you to cite. Run
skyquery citefor a ready-to-paste block. - Deterministic, not guessed. The correctness-critical logic (unit and frame conversion, cross-match, provenance) is pure and separated from I/O, and tested against known reference values.
- A good citizen by construction. Human-scale rate limits and an on-disk cache are hard-coded floors, so a free service is never hammered.
The tool surface
The MCP server exposes normalized, well-documented tools an assistant can chain (resolve a name, get its ephemeris, find its papers):
| Domain | Tools |
|---|---|
| Object intelligence | resolve_object, object_dossier |
| Ephemerides & small bodies | get_ephemeris, get_small_body, apophis_demo |
| Catalogs & cross-match | cone_search, crossmatch |
| Literature | search_literature |
| Analysis (no network) | convert_units, convert_frame, distance_from_parallax |
| Wonder layer | astronomy_picture_of_the_day |
| Provenance | session_citations |
The CLI mirrors these: resolve, ephemeris, small-body, cone, literature, convert,
frame, apod, demo, cite, plus setup, login, status. Every command takes --json for
machine output and --live to leave replay mode and query the real services.
For the auditor
You should be able to verify the trust model from readable code and passing tests:
git clone https://github.com/KarthikSubramanian07/skyquery && cd skyquery
uv pip install -e ".[dev]"
pytest # green with zero network and zero keys
The suite proves the three things that matter: numerical correctness (unit and frame conversions checked against astropy's own transforms and real captured reference values), good citizenship (the rate-limiter throttles bursts and backs off), and credential secrecy (keys touch only the keychain, and a test greps all log output to prove no token ever leaks). See DECISIONS.md for the architecture and the OSS ledger.
Built on the shoulders of
astropy · astroquery · astroplan · the Model Context Protocol SDK · and the free public services listed above. This tool is a thank-you note to all of them.
Support
SkyQuery is free forever and costs nothing to run. If it saved you a late-night tab-juggling session, you can buy me a coffee ☕.
License
MIT. Go build something with the sky.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.