Skip to main content

​
Data conventions

​
Response metadata

List endpoints return data, has_more, next_cursor, and a meta object. Depending on the endpoint, meta and the matching headers describe freshness and completeness: Check GET /v1/status (free, no key needed) for each dataset’s health (ok, degraded, stale, or unknown), its live lag and, where available, its complete_to. GET /v1/datasets lists every dataset with its coverage and endpoints, and GET /v1/changelog lists features, deprecations, data corrections, and incidents.

​
Pagination

List endpoints use cursors. Pass limit (each endpoint documents its default and maximum) and follow next_cursor until has_more is false.
  • A cursor is signed, bound to your workspace and the exact query, and valid for 24 hours. An expired or mismatched cursor returns 400 invalid_parameter.
  • Responses are capped at 2 MiB (16 MiB on batch endpoints). A request that would exceed the cap returns 400 range_too_large: lower limit and page through.
  • Request bodies can be at most 64 KiB (1 MiB on batch endpoints, 128 KiB on the priority lookup).

​
Response formats

List endpoints return JSON by default. Send Accept: text/csv for CSV, or Accept: application/x-ndjson (also application/jsonl) for newline-delimited JSON. Rows and price are the same in every format. Test keys always return JSON. In CSV and NDJSON, pagination moves to the X-Has-More and X-Next-Cursor headers, and the freshness headers above apply.

​
Async jobs

Async jobs run computations that take too long for a normal request. They are available on Growth, Scale, and Enterprise.
  1. POST /v1/jobs with an Idempotency-Key header and a body such as {"kind": "account_fills_summary", "params": {...}}.
  2. Poll GET /v1/jobs/{id} until the job is done. Asking for the result too early returns 409 job_not_ready.
  3. Read the result from GET /v1/jobs/{id}/result, paginated like any list.
The only job kind today is account_fills_summary. For one wallet over up to 90 days, it returns per coin: fill count, buy and sell notional, fees, realized PnL, and first and last fill. At most 2 jobs can run at a time and 100 per day. Results are kept for 24 hours. Creating, listing, polling, and cancelling jobs is free. Reading the result costs 10 credits plus 1 per row, and the first page also carries the job’s compute fee: 20 credits per upstream call plus 1 credit per 10 fills scanned.