I built Tork after watching AI agents get deployed into production with zero governance infrastructure — no audit trails, no PII controls, no way to prove compliance after the fact. The problem isn't the model. It's the harness. Once an agent has tool access, alignment guarantees mean very little without a middleware layer that can intercept, inspect, and enforce. Tork governs at that layer. Every API call through Tork generates a cryptographic compliance receipt — immutable proof of what the agent saw, what was redacted, and what decision was made. We run our own internal agent (NanoClaw) governed entirely by Tork, so every feature we ship is battle-tested against real governance requirements. What's live today: PII detection across 18 types with Luhn validation and semantic date checking, SHA256 exact-match caching, session-depth risk scoring, protocol governance for A2A, ACP, AG-UI and Consent, 11 SDKs, 116 framework integrations. Free tier at tork.network — 5,000 API calls/month, no credit card. Also wrote a free book on this: tork.network/book Happy to answer any questions about the architecture or governance philosophy.

Tork Network

The governance layer your AI agents are missing

The governance layer your AI agents are missing