Skip to content

Sovereign proof for every AI agent action.

Mint a capability. Gate the call. Verify the receipt — offline, in five languages, with no vendor contact.

Every tool call an agent makes on a gated path produces a capability receipt: what the agent was authorised to do, by whose authority, against a policy proof, and what action was executed. Signed under your organisation’s own key. Verifiable offline by any third party — a regulator, an auditor, a partner organisation — without contacting the vendor.

The soundness claims are mechanised in Lean 4 (three theorems, zero sorrys) and policy conformance is machine-checked per tool by an SMT prover, exactly as stated — and no more. Each page here ends with what is verified and what is assumed.