Public-data tiers#
| Tier | Monthly price | Requests/min | Requests/month | Rows/page | History | Keys |
|---|---|---|---|---|---|---|
| Anonymous | USD 0 | 60 | No configured cap | 100 | 400 days | 0 |
| Free key | USD 0 | 60 | 5000 | 500 | 730 days | 1 |
| Pro | USD 49 | 300 | 100000 | 5000 | Full retained archive | 3 |
| Team | Contact | 600 | No configured cap | 5000 | Full retained archive | 10 |
These are build-time configured values. Runtime overrides and the limits returned by your actual response take precedence. Team is the contact tier: it has no list price, and its limits above are the configured defaults for an agreement. History limits describe access to retained records, not a promise that every source goes back that far.
How usage is counted#
A keyed public-data request consumes one metered request during admission, before dataset query validation and retrieval. It is not counted per returned row. CSV is one page and one request, subject to the same page ceiling. Anonymous usage has no per-key monthly ledger. Inspect your own usage with GET /api/usage and a Bearer key.
Research and MCP#
The Pro research allowance is 5,000 combined research API and MCP tool calls per calendar month, separate from the public-data allowance. Read discovery for transport-specific throttles. MCP protocol initialization is not a metered tools/call.
When throttled#
Use response headers, especially Retry-After and the rate-limit fields. A monthly limit can require waiting until the next calendar reset, not a rapid retry loop. Never retry billing or signup blindly. Error guidance explains which failures are retryable.