The Data Intelligence Layer
of Bharat.
First-party Indian geospatial, civic, environment, POI and transit intelligence — keyed to H3 hexes, cached at the edge, and shaped for both humans and autonomous LLM agents. 30-second signup. No credit card.
# 1. Get a key
$ curl -X POST https://api.gridrock.ai/oauth/register \
-H 'Content-Type: application/json' \
-d '{"client_name":"my-bot","tos_accepted":true,"intended_use":"product","contact_email":"me@example.com"}'
# 2. Make a call (60 req/min as an agent)
$ curl https://free.gridrock.ai/v1/intel/hex/892a100c87ffff \
-H 'Authorization: Bearer agtok_…' \
-H 'Accept: application/vnd.gridrock.agent+json'Four suites. One substrate.
Each suite ships its own caching, attribution, and rate-limit profile, but they all share the same tier matrix, error shape, and agent response contract — so what you learn in one, you know in all.
Lat/lng ↔ H3 ↔ admin primitives. Reverse, forward, autocomplete, h3 — math + first-party Indian gazetteer.
H3 cell math, semantic labels, freshness state, polygon → cells coverage. Hex is the spine of the platform.
AQI, weather, 24h forecast — keyed to H3 cells, polled at signal-defined cadence by the GridRock fleet.
Authoritative ward, pincode and state polygons; transit stops (BMTC, BEST, Metro, Mono, Suburban); points of interest.
We don't tolerate agents.
We design for them.
Every endpoint exposes a token-efficient application/vnd.gridrock.agent+json shape (~42% smaller). Every error has a next_action enum so an LLM can recover without a human. Every route advertises its rough cost in LLM tokens via X-Cost-Per-Call-Tokens. Self-onboarding is one POST.
# Add to ~/.claude/claude_desktop_config.json
{
"mcpServers": {
"gridrock": {
"command": "npx",
"args": ["@gridrock/mcp-server"]
}
}
}
# First run handles RFC 7591 registration, persists creds,
# auto-refreshes the agtok_* JWT, and exposes 14 tools.Built for India, on India infrastructure.
Every byte stored in ap-south-1. Every cache POP that serves you sits in Bharat. Every dataset is licensed CC-BY-SA so the public good stays in the public good. We exist because Bharat's spatial intelligence shouldn't have to round-trip through California.
_metaWhy a layer, not just an API?
Bharat-grade decisions need Bharat-grade context: a Bandra cell knows it's in BMC ward H West, that the BEST 33-Ltd halts 90m away, that the CGWB well 1.2km north is rated saline, that the AQI was 142 at 3pm, that the 24-hour forecast says rain at 0700. We give you all of that in one HTTP call: /v1/intel/hex/<h3>. That's a layer.