Quick Verdict
Mireye Earth turns a US coordinate into structured geospatial fields for agents, attaching a source, upstream URL, dataset vintage, fetch time, and confidence rating to each value. It offers POST /v1/ask, deterministic POST /v1/fetch, a public field catalog, and a hosted MCP endpoint at https://api.mireye.com/mcp. The service is useful for underwriting screens, lending diligence, and energy or data-center siting workflows that need provenance instead of an LLM’s unsupported estimate.
Provenance is the product, but it does not replace professional GIS analysis or a final decision. Coverage is US-only. Sources include USGS, NOAA, USDA, FEMA, and other federal agencies, plus open or commercial sources such as Sentinel-2, Overture Maps, and Regrid, so describing every field as federal would be inaccurate. Confidence is a service-level quality signal, not regulatory approval. High-impact workflows must inspect vintage, notes, status, and partial failures, then verify critical facts against primary data and applicable rules.
Best For
- Agent developers needing traceable US geospatial context over MCP or HTTP
- Insurance and lending teams using hazards, terrain, wetlands, or parcel data for screening and human-review queues
- Site-selection analysts combining power, water, climate, roads, parcels, and natural hazards
- Auditable applications prepared to retain provenance, timestamps, confidence, request IDs, and failure records
- Not ideal for non-US locations, full GIS modeling, automatic underwriting or credit decisions, or teams unable to review high-impact output
Key Features
- Natural-language Q&A through
/v1/ask, returning citations andfields_used - Deterministic field access through
/v1/fetch, with values, units, source, vintage, timestamps, TTL, confidence, status, and notes - Public discovery catalog at
GET /v1/meta/fieldsfor live field, preset, type, source, and coverage metadata - Hosted and local MCP options, including
api.mireye.com/mcpand the localmireye-mcpadapter - Explicit failure semantics, distinguishing successful absence from a
partial_failuresentry inside an HTTP 200 response - US data layers spanning terrain, land cover, buildings, transport, utilities, energy, climate, hazards, parcels, and boundaries
Use Cases
- Screen property coordinates for flood, wildfire, terrain, wetland, and natural-hazard indicators
- Compare candidate energy or data-center sites before engineering and permitting diligence
- Support lending and title reviews with parcel, boundary, flood, and protected-land fields
- Batch-evaluate coordinates and route low-confidence, absent, or failed fields to human review
- Explore with
/v1/ask, capturefields_used, and replay those exact fields through/v1/fetch
Pricing
As of July 21, 2026, Mireye Earth was free during early access while coverage and rate limits were tuned with design partners. Post-early-access pricing had not been announced. The field catalog is public, while data calls and MCP operations require account authentication. Free access should not be interpreted as unlimited throughput or a production SLA; teams should confirm quotas, concurrency, support, caching, and commercial terms.
The local MCP adapter is published under MIT, which covers the adapter code. It does not automatically license the hosted service, API output, Mireye branding, or every upstream dataset under MIT. Federal public data, open datasets, and commercial sources can carry different attribution, caching, redistribution, and use conditions. Commercial products should inspect the service terms and each field’s source_url.
Pros
- Per-field source, URL, vintage, timestamp, and confidence support an inspectable agent evidence chain
/v1/asksupports discovery while/v1/fetchsupports repeatable, named-field calls- MCP, HTTP, and a public machine-readable catalog provide flexible integration paths
- Honest distinction between
absentandpartial_failuresis useful in production clients - Presets reduce the work required to coordinate many heterogeneous sources
Cons
- US-only coverage
- Early access has no stable public production pricing or SLA
- Aggregation inherits source freshness, resolution, gaps, and licensing differences
/v1/askfield selection may vary; controlled workflows need deterministic fetches- Citations and confidence do not create a legal, engineering, underwriting, or lending conclusion
Alternatives
| Tool | Better for | Strength | Limitation |
|---|---|---|---|
| Cyanheads NWS Weather MCP Server | Agents needing only US weather observations | Narrower and simpler scope | No parcel, terrain, energy, or multi-hazard catalog |
| Singapore Government Open Data MCP | MCP users needing Singapore public data | Direct regional government datasets | Different geography and cannot replace US coverage |
| Brave Search API | Agents needing current web research | Broad retrieval for narrative context | Search results are not standardized geospatial fields with field-level provenance |
| Tavily | Agents retrieving location-related news, policy pages, and web context | Agent-oriented search and page extraction | Adds web context only; it does not supply structured geospatial fields or replace Mireye’s field-level provenance and status semantics |
| Direct USGS/FEMA/NOAA integrations | GIS teams requiring full source control | Primary schemas and independent verification | Teams must build spatial joins, caching, normalization, and failure handling |
FAQ
What is Mireye Earth?
It is a US geospatial API and MCP service. Give it a coordinate and question or named fields, and it returns sourced data that agents can use as traceable context.
Does it use only US federal data?
No. Many core sources are federal, but official agent documentation also lists open and commercial sources including Sentinel-2, Overture Maps, and Regrid. Inspect each returned field’s source.
Can high-confidence results approve insurance or a loan automatically?
No. Confidence is a quality signal, not regulatory approval. Professional judgment, source documents, field verification, applicable rules, and organizational approval controls remain necessary.
Does HTTP 200 mean every field succeeded?
Not necessarily. /v1/fetch can return partial_failures inside a 200 response. Clients must inspect that array and distinguish failures from a valid status: absent field.
Is all Mireye data MIT because the MCP adapter is MIT?
No. MIT covers the adapter code, not automatically the hosted service, API output, or upstream datasets. Review Mireye terms and source-specific rights separately.
Is Mireye Earth free?
It is free during early access, with later pricing unannounced. Confirm rate limits, quotas, SLA, support, and future billing before production deployment.
Bottom Line
Mireye Earth is useful when a US geospatial agent needs field-level provenance rather than a language model’s guess. Evaluate it by fetching the public catalog, testing real coordinates, forcing absent and partial-failure cases, and independently checking primary sources. It belongs in production only when the team preserves provenance, handles failures, respects source rights, and keeps qualified human verification in every high-impact workflow.