
Noirlab Datalab MCP server
online · 100% uptimeNOIRLab Astro Data Lab — datalab.noirlab.edu.
not scanned yet
0 stars
Reviews
Write oneNobody has reviewed Noirlab Datalab yet.
If you have run it, two minutes of your experience saves the next person an afternoon.
Noirlab Datalab tools
Tool list not cached yet. `describe` through the gateway fetches it live.
How Noirlab Datalab compares
Against every graded MCP server, from our own scans and uptime checks. Updated after each scan run.
- Safety score
- 77/100
- higher than 50% of 48,255 servers
- Uptime
- 100%
- of our recent checks; the typical remote server answers 100%
- Response time
- 331 ms
- faster than 44% of 30,031 servers
Release history: 5 versions in the registry since 2026-09-25, latest 2026-11-03, typically one every 6 days.
Install Noirlab Datalab in Claude Code, Cursor or VS Code
claude mcp add --transport http noirlab-datalab https://gateway.pipeworx.io/noirlab-datalab/mcp
What the publisher says
From the Noirlab Datalab repository's README, as published. We do not edit it. Read it on GitHub
@pipeworx/noirlab-datalab
NSF NOIRLab's Astro Data Lab — the DECam, Mayall and Blanco survey catalogues (Legacy Surveys/DECaLS, DES, DELVE, DECaPS, SMASH, NSC) alongside mirrored all-sky catalogues (Gaia DR3, AllWISE, CatWISE, 2MASS, SDSS, DESI), queryable as database tables. Around 130 schemas and tens of billions of rows.
Part of Pipeworx — an MCP gateway connecting AI agents to 1679+ live data sources.
Tools
or one table's columns. Call it first: table names are survey-specific and a guessed name is a query error, not an empty result.
- datalab_tables(schema?, table?, limit?) — list the survey data releases, one release's tables,
one, so a bare query cannot scan a billion-row table unbounded.
- datalab_query(sql, limit?) — run a read-only ADQL SELECT. A TOP n is injected when you omit
sky position, nearest first, with exact angular separation in degrees.
- datalabconesearch(table, ra, dec, radius?, columns?, limit?) — sources within a radius of a
Auth
Keyless — but only through the IVOA TAP endpoint. See the first trap below.
Data sources
Anonymous, and the only door this pack uses.
- — REQUEST=doQuery&LANG=ADQL&FORMAT=csv&QUERY=….
Traps
datalab Python client uses and what the docs lead with, but without a login it answers 401 The provided security token is invalid — and the widely-quoted X-DL-AuthToken: anonymous.0.0.anonaccess answers HTTP 200** with the body Error in query(). Try checking that you are logged in. A 200 carrying a failure is the worst shape a failure can take, and it is why this pack speaks TAP only. Measured 2026-09-17.
- /query/query?sql= is NOT anonymous, despite being the documented front door. It is what the
200 with a VOTable whose single INFO element carries QUERY_STATUS="ERROR". Read as data that is an empty result set. Every response is checked for that element before it is parsed.
- TAP reports every error as HTTP 200. Syntax errors, unknown tables, timeouts: all come back
CONTAINS(POINT('ICRS',ra,dec), CIRCLE('ICRS',…)) parses and is handed to PostgreSQL verbatim, which answers function circle(numeric, numeric, numeric) does not exist. The Q3C functions the backend actually has (q3cradialquery) are rejected by the ADQL parser one layer up. So both obvious cone-search spellings fail. datalabconesearch builds the search from trigonometry both layers accept: an indexable RA/Dec box (RA widened by 1/cos(dec), wrapping at 0h) to narrow the scan, then an exact great-circle distance to trim the box corners off and order by true separation.
- ADQL geometry is not implemented, and neither is Q3C.
Galactic plane selectively, so a northern target legitimately returns nothing from lsdr10 or desdr2. The zero-result note says to re-check against gaiadr3.gaiasource, which is all-sky.
- Row limits are SELECT TOP n, not LIMIT n. ADQL, not PostgreSQL.
- Zero rows is usually footprint, not absence. The DECam surveys cover the southern sky and the
Quick Start
Add to your MCP client (Claude Desktop, Cursor, Windsurf, etc.):
{
"mcpServers": {
"noirlab-datalab": {
"url": "https://gateway.pipeworx.io/noirlab-datalab/mcp"
}
}
}What this endpoint actually serves
tools/list at https://gateway.pipeworx.io/noirlab-datalab/mcp returns the tools in the table above plus the shared Pipeworx meta-tools — askpipeworx, discovertools, search_within, remember/recall and the rest of the gateway-wide set. So the tool count you see is larger than this table: a single-pack endpoint currently lists roughly 30 shared tools alongside the pack's own. The connection's initialize response states its exact scope, and is the authoritative answer for a given day.
Shortened. The full README is on GitHub.
Nothing above is checked by us. What we check is on the safety report.
Noirlab Datalab: common questions
- Is Noirlab Datalab MCP server safe?
- Mostly: it is graded B (77/100). Read the Noirlab Datalab safety report
- How do I install Noirlab Datalab?
- It runs remotely at gateway.pipeworx.io. Add it to Claude Code, Cursor, VS Code or Claude Desktop with the snippets above, or call it through the mcp.market gateway without installing anything.
- Does Noirlab Datalab need an API key?
- Yes. The registry entry asks for
PIPEWORX_API_KEY. - Is Noirlab Datalab maintained?
- The last commit was 40 days ago (2026-09-25). The latest release is v0.2.1.
- Is Noirlab Datalab up?
- 100% of our last 28 checks got an answer. We check remote servers about four times a day.
- How does Noirlab Datalab compare with other MCP servers?
- safety score 77/100 (higher than 50% of 48,255 servers); uptime 100% (of our recent checks; the typical remote server answers 100%); response time 331 ms (faster than 44% of 30,031 servers). Full comparison.