Mireye (YC S26): The Physical World Operating System for AI Agents

Mireye provides the essential infrastructure required for AI agents to interact with, understand, and make decisions about the physical environment. By offering a single, unified API, Mireye delivers critical data, enrichment, signals, and tools that bridge the gap between LLMs and real-world geography.

---

🌍 The Problem: AI's "Physical Blindness"

Standard AI models struggle with precise spatial data. They often hallucinate or provide vague approximations when asked about specific physical coordinates.

Scenario: What is the elevation at $40.7128, -74.006$?

Standard Claude: "I don't have real-time access to elevation databases." "Based on general knowledge, Manhattan elevations range from sea level to about 10-50 meters... would you like some online tools?"

The Mireye Solution: When powered by Mireye, the same agent provides a surgical answer:

Claude + Mireye: "The elevation at $40.7128, -74.006$ is $\text{13.03 meters}$ above sea level (NAVD88 datum)."

elevation = 13.03m | source = USGS_3DEP_COG | fetched = 2026-07-28 | confidence = medium

---

⚙️ How It Works: The Infrastructure Loop

Mireye acts as the connective tissue between advanced AI agents (such as GPT-5, Claude, Gemini, DeepSeek, or Cursor) and a massive array of physical world data sources.

What the system delivers to the agent:

• Cited Data: Every fact is backed by a source.
• Enrichment: Contextual layers added to raw coordinates.
• Proprietary Fields: Specialized data not found in public catalogs.
• Signals & Tools: Actionable triggers for decision-making.
• Dynamic Expansion: If a field is missing, it is requested, built, and then available for all future users.

---

🏢 Industry Applications

Mireye transforms the question "Should I care about this specific location, and why?" into a programmable API call.

---

🛠️ The API Reference

Mireye provides a suite of POST endpoints designed for seamless integration via MCP (Model Context Protocol).

01 POST /v1/ask

The "catch-all" endpoint. Resolve addresses, fetch specific fields, and calculate distances in one go. It returns the execution plan for full replayability.

02 POST /v1/geocode

Converts a US street address into precise coordinates.

• Nuance: It specifies how the coordinate was found (e.g., interpolated vs. parcel-based).
• Strictness: Low-confidence matches are refused rather than guessed.

03 POST /v1/lookup

A comprehensive "stack" call. Returns:

• County, Tract, and Congressional District.
• Metro area and Timezone.
• Elevation and Flood status.

04 POST /v1/fetch

Retrieves typed values for specific fields or presets.

• Batching: Supports hundreds of locations per call; individual address failures do not crash the request.
• Presets include:
  • flood_risk (13 fields)
  • solar_siting (26 fields)
  • wildfire_underwrite (6 fields)
  • grid_interconnect (29 fields)

05 POST /v1/proximity

Calculates what raw data cannot. It offers four drive-time operations for the US and Canada:

1. Point-to-point distance.
2. Nearest-by-road (Airports, Rail, Ports, etc.).
3. Drive-time cutoff screening.
4. Labor shed analysis.

06 POST /v1/field-requests

The growth engine. Users can describe a needed data field in plain English. Mireye then attempts to match it to the catalog or queues a new build to integrate that data for everyone.

{
  "request": "Average annual snowfall for the last 10 years",
  "locations": [[40.7128, -74.006]],
  "priority": "high"
}