The archive · AI & Models · Product decision · 2026
Mireye bets AI agents need cited physical-world data; YC S26 API launch
Ansh Chokshi killed his site-screening app and built Mireye: one API giving agents cited US location data, tools and on-demand indexing.
Mireye
What the business is
Mireye sells an API and MCP server that give AI agents grounded US location data — 300+ fields across land, buildings, people, rules and risks, plus tools and change signals — with every value carrying its source, timestamp and confidence.
How it started
Ansh Chokshi, who sold his first startup Poker Pit at 21 after growing it from zero to $250K ARR in four months and later led AI/ML engineering at Seismic, hit the wall while building construction agents: they reasoned well about anything online but knew nothing about the ground underneath a site. A Fortune 500 insurer told him its engineers had given up on underwriting agents for the same reason — frontier models hallucinate when asked a specific question about a specific place. His first product, a niche site-screening app, tested well on places customers knew, but nobody cared about the app; they wanted the engine underneath.
What happened
Usage showed the signal: 311 of 317 catalog fields were queried with no dominant use case, so Chokshi killed the app and built Mireye as infrastructure — one API and MCP server covering 300+ US location fields growing about 20% week over week, with vertical presets for data-center siting, solar, flood risk and site selection. The distinguishing choices are typed absence and on-demand indexing: ask for a field not in the catalog and a long-running agent sources, tests and indexes it, usually within a day, after which it serves every future caller. First customers built agents that source off-market land deals for data centers about 100x faster than before.
No ending yet — it is still running.
Background
Mireye is a YC S26 startup selling the data layer for AI agents that act on the physical world: one API and MCP server that turns a US address or coordinate into cited, provenance-rich facts — owner, acreage, structures, flood risk, nearby power — plus deterministic tools and change signals such as a rezoning filing.
Founder Ansh Chokshi came from construction agents, where he found that models reason well about anything online but hallucinate when asked a specific question about a specific place; a Fortune 500 insurer told him its engineers had given up on underwriting agents for exactly that reason. His first product, a niche site-screening app, earned correct answers but no customers — usage showed 311 of 317 catalog fields queried with no dominant use case, so he killed the app and built the engine people kept asking for.
Mireye's catalog spans 300+ fields across land, location, people, rules and risks, growing about 20% week over week, with vertical presets such as data-center siting (90 fields), solar siting and flood risk. Two design choices define the product: typed absence — every field returns ok, absent or failed rather than a bare null that a model would fill with a plausible guess — and on-demand indexing, where a request for a missing field triggers a long-running agent that sources, tests and indexes it, usually within a day.
As of its September 2026 launch, Mireye reports first customers who built agents that source off-market land deals for data centers about 100x faster, with insurance, proptech, robotics and drone-planning use cases emerging. The company has not disclosed revenue or named customers, and its claim to be the Stripe-for-place-data will be tested by whether it can keep thousands of county-level sources fresh and trusted.
What has to be true
- Frontier models confidently guess at elevation, flood zones and parcel boundaries, and a wrong answer about a physical place is not a bug but a liability.
- Real-world data is fragmented across thousands of county and federal sources built for human PDF-and-GIS workflows, so an agent-ready index is a genuine missing layer.
- The typed-absence design turns 'we do not know' into a feature: customers say the refusals are why they trust the API.
- On-demand indexing inverts the normal data-company tradeoff — the catalog grows with every customer request instead of being fixed at launch.
What can be applied
An app nobody wanted revealed a market everyone needed: when every customer asks for the engine beneath your demo, the demo was the discovery vehicle — kill it and sell what they asked for.
Aftermath
As of 2026-09-03 Mireye is live with a shipped API and MCP server, a Launch YC listing from 2026-08-21, a Launch HN on 2026-09-03, and a 300+ field US catalog growing about 20% week over week. First customers report sourcing off-market data-center land deals about 100x faster. It has not disclosed revenue, funding beyond YC, or named customers; the open questions are whether per-field freshness across thousands of county sources holds at scale and whether the 'engine under every physical-world agent' bet becomes a durable franchise.
Sources
- Launch YC: Mireye: The easiest way to build agentic applications for the physical world
- Mireye (YC S26): la API que da contexto físico a los agentes de IA
- Launch HN: Mireye (YC S26) – Infrastructure for Physical World AI Agents
spotted an error? The archive wants to know.
Your turn
You just read one. Describe what you are building, and see who is betting on the same thing.
Free account · 3 free questions · no card