seiche.info

Signed delivery receipts for agentic commerce: submit what a paid service returned (plus its x402 settlement receipt and advertised schema) and receive an offline-verifiable Ed25519 attestation binding payment to delivery to grounded content — schema conformance, machine-verified claim verdicts, and a neutral accountability trail for disputes between agents and the services they pay.

Attested instrument-identity enrichment for AI agents: map a ticker, ISIN, CUSIP, SEDOL, FIGI, or instrument name to canonical FIGI records (name, security type, market sector, exchange) via Bloomberg open symbology, with provenance (source, retrieval time) attached to every answer. Built for agents that must know WHICH security a claim, order, or document is about before acting on it.

Claim grounding and citation attestation for AI agents: verified enrichment that adds a veracity field to text. Each factual claim is grounded against live web sources (Wikipedia + world news) and returned with a machine-verified verdict (supported/refuted/unverified), a confidence score, and cited sources. Refuses to guess on conflicting evidence. Built for research agents that must cite, citation-locked content, and compliance checks.

Claim extraction for agent verification loops: split text into independently checkable atomic factual claims via auditable rule-based decomposition (no LLM guessing), with a signed receipt bound to the input hash. The cheap first step of the loop — extract here, then ground what matters via /verify or /check.