Sanna Information

Sanna is trust infrastructure for AI agents, built on open-source protocol infrastructure with two equal pillars: constitution enforcement and cryptographic receipts. Constitution enforcement blocks out-of-scope agent actions before execution, not after. YAML constitutions define boundaries, invariants, authority rules, halt conditions, and escalation targets. Every agent action passes through Sanna's governance layer before it can execute. Cryptographic receipts provide tamper-evident, Ed25519-signed proof of what governance was applied at the moment of action. Every receipt binds the constitution hash, proving which specific approved policy version was governing an agent when it acted. Receipts are independently verifiable offline by any third party without platform access. Enforcement without proof is incomplete. Proof without enforcement is audit theater. Both must be present in every deployment. Sanna governs every surface an agent touches: MCP tool servers, CLI subprocess execution, and HTTP API calls. One constitution governs all surfaces. Same receipt format, same verification, same proof.Open-source SDKs ship in Python and TypeScript. The Sanna Protocol specification defines receipt format, canonicalization (RFC 8785), signing, and verification as a portable, vendor-neutral standard. Sanna Cloud is the hosted governance platform built on AWS, layering fleet dashboards, managed gateways, encrypted constitution storage, drift detection, alerting, and SIEM integration on top of the open-source foundation. Designed for teams and enterprises operating agents in regulated or high-trust environments, with a path from single-agent governance through full enterprise fleet management. Model providers cannot credibly govern themselves. Sanna exists because governance infrastructure requires independent verification, and independent verification cannot itself be a black box.