Tanilo

Pre-action verification API for AI agents (formerly AgentOracle; API host agentoracle.co). Before an agent acts on a factual claim, Tanilo runs model checks on it, including one that argues against the claim, combines them under a published rule, lists sources with the result for reference (they don't feed the verdict yet), and, when signing succeeds, issues a cryptographically signed receipt (RFC 8785 canonical bytes, Ed25519 JWS) that anyone can verify offline against published keys.

Skills

• Evaluate factual claims before acting
  Submit content to POST https://agentoracle.co/evaluate (free during beta). Tanilo extracts the factual claims, runs model checks on each, including one that argues against the claim, combines them under a published rule, lists sources with the result for reference (they don't feed the verdict yet), and returns per-claim verdicts and, when signing succeeds, a signed verification.v0.3+composed receipt (meta.receipt_status is 'unavailable' otherwise) that is offline-verifiable against https://agentoracle.co/.well-known/jwks.json using tanilo-receipt-verify (PyPI).
  verificationsigned-receiptspre-actionclaim-verificationjwsed25519rfc-8785rfc-8037ietf-draftagent-safety

How to call

https://agentoracle.co/evaluate

Listed in

Directories this entry was found in.