Coverage
Both venues are read as public data. No account, wallet or identity is recorded with a trade.
Categories, with the number of events in the latest graph on 17 September 2026:
GET /v1/categories returns the current counts.
Volumes
Observed on 17 September 2026 fromGET /v1/status and GET /v1/books/health:
Retention
What is not provided
- No trading. There is no order placement, no positions, no balances and no venue credentials. Every endpoint is
GET; the MCP server is read-only. - No accounts. No endpoint creates, lists or revokes keys, and nothing is stored about you beyond request counts.
- No identities. Trades carry the venue’s trade id, price, quantity and time, never the account or wallet behind them.
- No raw order-book export. Raw level changes are kept for 7 days and dropped; their redistribution rights are under review and until that clears they are not exported or served in bulk. Snapshots and top-of-book history are kept.
- No proofs from prose. Markets without structured terms are listed, priced and searchable, but no relation is proven for them.
- No quotes. Prices are observations recorded on a crawl. They can move before an order reaches the venue.
- No predictions. A relation says how two contracts’ outcomes constrain each other, not how either will resolve. Calibration and strategies describe settled markets; past gaps can close.
- No browser access to the stream.
EventSourcecannot send a key header; connect from a server.
Informational leads
A lead is offered asexecutable only when its live order books could fill at least one whole pair for less than it pays after fees and minimum sizes. One class of lead is never offered as executable however the books look: cross-venue relations between economic-release markets (for example CPI or a Fed decision) where the two venues settle a delayed or cancelled release differently. Such a lead carries depth.limited_by = "informational" and the issue:
executable=true lists and of lead.executable webhooks.
Fees
estimated_fee_usd on a lead and fees_usd in its depth are estimates from published schedules. Where the crawl has read a market’s own fee terms, they are used; otherwise the taker fee is 0.07 × p × (1 − p) per contract, Kalshi’s published coefficient. Venues change fees; check the venue before trading.