The free tier needs no key and no signup: 1,000 calls a month, 200 of which reach a venue upstream. Every page on this site is one of these calls.
| Endpoint | Answers |
|---|---|
| /v1/size | can it take my size, in plain words? |
| /v1/markets/search | which market did I mean? |
| /v1/equivalence | are these two the same bet? |
| /v1/mcp-equivalence | the same, by identifier |
| /v1/preflight | will this order work? |
| /v1/rest | rest or cross, and how long? |
| /v1/copyability | is this wallet worth copying? |
| /v1/venues/conventions | how does this venue lie? |
| /v1/evidence | a citable, reproducible verdict |
| /v1/watch | tell me when the rules change |
| /v1/scorecard | were we right before? |
Each description leads with the failure it prevents, because an agent does not know it has a problem until something tells it. A model reading two rule sets will report they are the same bet; it has no way to see that one admits voluntary guidelines and the other is silent.
claude mcp add --scope user clause -- python -m clause.mcp.server
identify_market takes plain words and refuses to choose between contracts that are different bets.
ClauseError — your request was wrong, and the message says what to send instead.
ClauseUnavailable — the check did not run. Never treat this as a pass.
Neither inherits from the other, so catching a malformed-request error cannot silently swallow an outage.
The same rule shapes the verdicts: INSUFFICIENT_DATA is distinct from LIQUID and never renders as a pass.
Getting this wrong is the most common first error.
| Name | Looks like | Where from |
|---|---|---|
| market_key | 0xcdbf64cb… | Polymarket conditionId, from /v1/markets/search |
| market_key | KXAIBILL-26JUN-27JAN01 | Kalshi ticker, from /v1/markets/search |
| token_id | long decimal string | /v1/markets/{condition_id}/tokens |
A Polymarket slug is none of these and will 404.