Mireye
An API that gives AI agents US physical-world data: cited place data, fields, parcel information, and signals in a single call.
About Mireye
Mireye is an infrastructure layer for AI agents in the physical world: Claude, ChatGPT, Kimi, Gemini, Cursor, Grok, Copilot, DeepSeek, or custom agents ask Mireye questions through a single API and MCP tools, and Mireye returns cited data, enrichments, proprietary fields, and signals — a data index of the physical world. It is not a map, a GIS tool, or a data reseller; instead, it provides four categories of capabilities under a shared auth, catalog, and provenance model, with data drawn from public agencies such as USGS, NOAA, EPA, and EIA, plus open data and clearly labeled licensed sources, all normalized to one schema. Every field comes with a source name, source URL, retrieval timestamp, and confidence, so agents can pass provenance straight to end users or auditors. Coverage is limited to the United States: /v1/ask, /v1/fetch, and their MCP tools serve only US locations, with primary bounds of latitude 18–72 and longitude -180 to -65, plus an additional range covering the western Aleutian Islands; out-of-bounds requests return a 400 coord_out_of_bounds error.
Best for
Best use cases
Insurance underwriting: one of the use cases listed on the official comparison page (e.g., an insurance underwriting question in Miami).
Construction site risk assessment: one of the use cases listed on the official comparison page (e.g., a site risk in central Iowa).
Defense and govtech: one of the use cases listed on the official comparison page (e.g., Arlington).
Renewable energy site selection: one of the use cases listed on the official comparison page (e.g., rural Kansas).
Mortgage and title: one of the use cases listed on the official comparison page (e.g., Lower Manhattan, New York).
Compare before choosing
According to the official comparison page: asking the same question about the same US coordinate to Mireye /ask and to a frontier model with web search, Mireye streams back an answer computed from real data with citations, while the baseline model can only access the open web.
Use cases listed on the official comparison page include insurance underwriting (Miami), construction site risk (central Iowa), defense and govtech (Arlington), renewable energy site selection (rural Kansas), and mortgage and title (Lower Manhattan, New York).
Features
One API plus MCP tools
Claude, ChatGPT, Kimi, Gemini, Cursor, Grok, Copilot, DeepSeek, or custom agents can call Mireye through one API and MCP tools.
Citations on every field
Every field comes with a source name, source URL, retrieval timestamp, and confidence, so agents can pass provenance directly to end users or auditors.
Authoritative data sources
Data comes from 96 authoritative sources, primarily federal datasets, along with county-level records and open datasets.
/ask natural-language queries
Ask natural-language questions about any location in the US, with answers computed from cited fields.
Field fetch
327 cited, typed place-fact fields; geocoding flags precision and rejects coarse matches instead of guessing.
Parcel lookup
Resolve an address or coordinates to a standardized parcel, parcel ID, boundary geometry, and registered owner, with county and census tract codes; billing only on successful matches.
Custom field requests
Fields not yet in the catalog can be requested; they are first matched against the catalog, and confirmed gaps enter a queued build, after which they become available to all callers.
Pros
- One API and one MCP server provide cited geospatial data, enrichments, and proprietary fields, with each field carrying source, URL, timestamp, and confidence.
- Data comes from 96 authoritative sources, primarily federal datasets, along with county-level records and open datasets.
- Credits are a single pricing unit across all endpoints, and the credit cost of each call type is published on the pricing page.
- No plan charges beyond your quota — once credits run out, usage simply stops.
Cons
- US coverage only: /v1/ask, /v1/fetch and their MCP tools support only coordinates and addresses within the United States; out-of-bounds requests return an error.
- Different data layers update at different cadences: for example, elevation is largely static, FEMA flood maps update on FEMA's schedule, and USDA cropland data updates annually — data freshness must be checked per use case.
- Geocoding rejects coarse matches outright instead of guessing, so loosely matched address queries may return no results.
Pricing Plans
Free
- 5,000 credits per month
- 1 field request at signup
- 20 requests per minute
- Parcel lookup: 300 credits
- No overage billing
- Community support
Build
- 25,000 credits per month
Enterprise
- Custom SLAs and SLOs
- Dedicated market coverage
- Priority support with direct access to the team
- Usage-based pricing
* Pricing information is for reference only. Please refer to the latest announcement on Mireye's official website.