A read-only data connector from Avanti Fellows, for Claude.
Ask questions in plain English about Indian higher and school education — admission cutoffs, college directories, enrolment, accreditation, seat capacity — and Claude writes and runs the SQL for you. No BigQuery account, nothing to install, no credentials to manage.
Do this on a computer — claude.ai in a browser, or the Claude desktop app. Connectors are set up there, not in the phone app. It is a one-time setup on your account; afterwards you just use Claude normally.
In Claude, go to Settings → Connectors and choose + Add custom connector.
Paste the address above. Leave everything under Advanced settings empty — the server registers Claude automatically, so there is no client ID or secret to enter. Click Add.
Click Connect, then sign in with Google and approve. Open a new chat and ask “what data can you see?” — Claude will introduce itself and list what it has.
On a Team or Enterprise plan? An owner adds it once under Organization settings → Connectors (Add → Custom → Web). Everyone else then does step 3 only.
Using Claude Code? One command replaces all three steps:
claude mcp add --transport http india-education-data https://india-education-data.avantifellows.org/mcp
| Admissions | JoSAA, NEET, KCET, MHT-CET, TNEA, WBJEE, KEAM, OJEE, GUJCET, AP & TG EAPCET, CLAT closing ranks |
| Institutions | AISHE directories, NAAC accreditation, NMC MBBS seats, NIRF rankings and first-party DCS figures |
| Schools | UDISE+ enrolment, teachers, facilities; NAS proficiency; MoE board results |
| Households | PLFS, HCES and CMS-E survey microdata |
| Avanti's work | Published analyses — the write-ups, methods and caveats behind the numbers |
This connector is scoped to public education data and Avanti's published analyses. It has no access to any student, donor or internal Avanti record — not by policy but by construction: it runs as a separate identity that can read one BigQuery dataset and nothing else, and the tools that reach anything else are not loaded into it.
Every query is read-only and rate-limited per person. Usage is logged against the signed-in account.
Anthropic's own guides cover the connector screens in general: about custom connectors and installing and using connectors. Anything specific to this connector — a refused sign-in, a table you expected to find — is ours, not theirs.