llms.txt

# Sanna — Trust Infrastructure for AI Agents

> Sanna is open-source AI agent governance infrastructure. It enforces agent
> constitutions at execution time, halts forbidden actions before they reach
> downstream systems, and generates Ed25519-signed Governance Receipts:
> tamper-evident, offline-verifiable proof that governance was applied at the
> moment of action. Two equal pillars — constitution enforcement and
> cryptographic governance receipts. Python and TypeScript SDKs, vendor-neutral,
> AGPL-3.0. Built by Nic Allen (15 years in regulated trust systems: pharma
> GxP/FDA, Google Cloud, cybersecurity).

## Core pages
- [Homepage](https://sanna.dev/): Product overview, the two pillars, deployment modes (Python library, TypeScript library, MCP gateway, OpenClaw plugin), and use cases for financial services, healthcare, and enterprise AI teams.
- [Papers](https://sanna.dev/papers/): Technical white papers and original research on agent governance and trust infrastructure.
- [Blog](https://sanna.dev/blog/): Analysis on AI agent governance, compliance, and audit evidence.

## Papers (original research)
- [Governance Envelopes for Verifiable Agent Handoff](https://sanna.dev/papers/governance-envelopes.html) (SANNA-WP-001, ~14,000 words; [PDF](https://sanna.dev/papers/governance-envelopes.pdf)): A portable, signed evidence graph for agent context, memory, citation, handoff, and action — plus a Receipt-Gated Handoff pattern that downstream agents can verify under local trust policy. Covers cross-vendor agent handoff and chain-of-custody review.

## Key articles
- [How to Govern AI Agents in Production](https://sanna.dev/blog/how-to-govern-ai-agents-in-production.html): Why enforcement plus proof is the complete governance layer; constitution YAML and signed receipt examples.
- [The Gap Between Policy and Proof in AI Agent Compliance](https://sanna.dev/blog/the-gap-between-policy-and-proof.html): Two audit firms independently identify the same gap — enterprises can't prove governance was applied at the moment of execution. Maps to SOC 2, NIST AI RMF, ISO 42001, EU AI Act.
- [Salesforce Headless 360 Creates a Governance Gap Agentforce Cannot Close](https://sanna.dev/blog/headless-360-governance-gap.html): Why headless SaaS removes the human UI checkpoint that compliance controls depended on, and why no platform vendor can close the gap from inside its own platform.
- [The Governance Gap Inside Claude Code](https://sanna.dev/blog/the-governance-gap-inside-claude-code.html): The Claude Code source leak reveals sophisticated runtime permission enforcement with no durable proof — the industry-default gap.
- [Live Red-Team: Constitutional Enforcement](https://sanna.dev/blog/live-demo-constitutional-enforcement.html): A multi-round, multi-model red-team exercise — 20+ attack vectors, signed receipts as ground truth, soft guardrails vs architectural enforcement.

## Code and specification
- [GitHub organization](https://github.com/sanna-ai)
- [sanna (Python SDK)](https://github.com/sanna-ai/sanna)
- [sanna-ts (TypeScript SDK)](https://github.com/sanna-ai/sanna-ts)
- [sanna-protocol (open specification, Apache 2.0)](https://github.com/sanna-ai/sanna-protocol)
- [PyPI](https://pypi.org/project/sanna/)

## Contact
- Email: nic@sanna.dev
- LinkedIn: https://www.linkedin.com/in/nicallen27