Conviction in.
Verified position out.
Klyr is the execution gate for autonomous market conviction. It turns an unstructured thesis into a structured claim, finds the market that actually represents it, verifies how that market resolves, checks the action against policy, and only then permits execution.
A thesis can be correct while the selected market is wrong.
An agent can be right about the underlying event and still select the wrong financial instrument — the wrong deadline, the wrong threshold, the wrong resolution source, a market that only covers one specific occurrence instead of the window the thesis actually describes. Klyr exists to catch that translation failure before it becomes an irreversible position.
The model may propose the position.
Klyr decides whether the position is valid.
Klyr separates probabilistic interpretation from deterministic enforcement. An AI model may parse a thesis, rank candidate markets, and extract resolution conditions — but it never declares a market verified, decides that a mismatch is safe, or approves execution on its own. When required information is missing, ambiguous, or contradictory, Klyr fails closed: BLOCK or ESCALATE, never a silent guess.
See it verify a thesis
This calls Klyr’s live ASP endpoints directly — the same routes an OKX.AI agent calls. Try the ambiguous example to see Klyr stop rather than guess.
A trader sees the original thesis, the structured interpretation, every rejected near-miss, the selected market, the resolution rules, and the full verification trail — before anything executes.
An autonomous agent submits structured input and receives a structured decision: PASS, BLOCK, or ESCALATE, with the full verification snapshot needed to act — or to stop.
Read the agent integration guide →Every decision is a receipt
Built to be called, not clicked.
Klyr is machine-callable infrastructure — an OKX.AI ASP with a documented API contract and an x402 payment gate on execution.