Evidence infrastructure for research agents

Messy records in. Research-ready business changes out.

Find unknown openings, permits, remodels, ownership changes, closures, and tenant transitions—resolved to real businesses and exact premises, with dated source evidence and honest coverage.

Integrate the tools Read llms.txt

Migration-ready events: 521
Source records: 1,139
Initial sources: 9

Agent retrieval / compact → evidence

user_request

“Find restaurant opening leads in Orlando observed this month.”

typed_tool_callfind_business_change_leads

{
  "jurisdictions": [
    "US-FL"
  ],
  "geography": {
    "cities": [
      "Orlando"
    ]
  },
  "intents": [
    "opening_leads"
  ],
  "industries": [
    "restaurant"
  ],
  "observedAfter": "2026-07-01T00:00:00.000Z",
  "projection": "compact",
  "limit": 25
}

structured_result

event + identity + confidence + coverage

Expand primary evidence only for the leads worth verifying.

API live

The v1 PostgreSQL data plane is public with 521 evidence-backed events and five HTTP + MCP tools. Payment enforcement is staged pending the recipient wallet.

Open production API

One contract, every channel

Five tools built around agent economics.

Search compact leads, pay to expand only selected evidence, monitor with stable cursors, and inspect coverage for free. HTTP and MCP share one schema.

HTTP + MCP$0.10 proposed

Find business-change leads

Discover previously unknown businesses and locations by geography, intent, event time, observation time, evidence, and confidence.

find_business_change_leads

Compact keyset-paginated events
Separate event and discovery time
Coverage returned with results

HTTP + MCP$0.02 proposed

Expand primary-source evidence

Expand only selected event IDs into immutable, dated source records with native identifiers and URLs.

get_business_event_evidence

Batch up to 50 event IDs
Content-addressed provenance
Built for citation and verification

HTTP + MCP$0.03 proposed

Research a location timeline

Retrieve chronological business changes for an exact premises or an explicitly requested parent building.

get_location_timeline

Suite identity preserved
Exact premises by default
Chronological event cursor

HTTP + MCP$0.05 proposed

Poll newly observed changes

Consume a stable observed-time delta feed, including corrections, supersessions, and retractions.

get_changes_since

Monitoring-safe cursor
Observed-time ordering
State corrections retained

HTTP + MCPFree

Inspect source coverage

Check supported sources, jurisdictions, collection windows, freshness, failures, and limitations before interpreting results.

get_business_change_coverage

No-match context
Per-source watermarks
Complete, partial, failed, stale

Intent-level discovery

Agents ask for jobs, not agency form codes.

Stable research intents sit above source-specific observation types, so new states and records expand coverage without breaking the way agents call us.

opening_leads

Opening leads

Early regulatory and location signals for establishments that may be preparing to open.

construction_activity

Construction activity

Permit, plan-review, build-out, expansion, and construction activity at business locations.

remodels

Remodels

Signals that an existing establishment may be rebuilding, remodeling, or expanding.

operator_changes

Operator changes

Ownership, operator, brand, or tenant transitions tied to a location over time.

closure_risk

Closure research

Authoritative status, license, inspection, or tenant-transition signals worth investigating.

compliance_activity

Compliance activity

License, inspection, enforcement, and regulatory activity for a business or location.

The differentiated layer

Not aggregation. Evidence infrastructure.

canonical_event

Opening signal at an exact premises

strong evidence

Effective time: When the permit or license event happened
Observed time: When the platform first discovered it
Identity: Brand + establishment + Suite 115, separately resolved
Public status: null — an opening signal is not proof of opening

Immutable receipts

Content-addressed snapshots and versioned source records make every event reproducible.

Reversible identity

Every match has a score, reason, algorithm version, and correction path.

Explicit absence

Coverage windows, failures, and limitations travel with the answer.

Distribution architecture

Direct first. Registries amplify.

One remote MCP and OpenAPI surface can be discovered through the official MCP Registry, Smithery, OpenAI Plugins, Cursor, Claude, Monid, BlockRun, and x402 Bazaar—without creating eight versions of the product.

See the channel plan

Commercial quality gate

Sell the supported claim—and expose its limits.

A 130-record stratified audit resolved every regulatory classification to its raw official source chain, with zero contradictions and zero claims that regulatory activity proved a public opening. Independent corroboration remains a separate dimension, never an implied fact.

130 / 130

Source chains verified

Every sampled classification resolved to the official records required by its rule.

445

Conservative default cohort

Paid search defaults to strong evidence and at least 0.75 classification confidence.

Opening overclaims

The product sells regulatory change signals, not unsupported claims that a door is open.

For research agents and builders

Start compact. Expand evidence. Keep the caveats.

Open integration guide Inspect OpenAPI