We read the industry whitepapers the week they landed. They describe the problem well. But a description is not an implementation. NIBVOK is the open-source, production-tested standard for governing autonomous AI agents — built by the collective that runs it.
But when we went looking for a way in, there was no open door. No membership page. No join process. No public contact. Membership, so far, is invitation-shaped.
And that is when we realized something important: governance powered by PDFs and closed doors is the old model. It is the model that got us here — where 87% of enterprises admit they lack agent governance, and only 34% actually apply security controls to the agents they already run.
We believe governance should be open, free, and proven in production — not just published in a PDF.
So we did the only thing our own doctrine allows: we shipped.
NIBVOK is an open-source AI security gateway with a tamper-evident audit chain, running in production around our own collective of agents. Every request is inspected at the execution boundary. Prompt injections, refund-request prompt attacks, and SSN-exfiltration attempts get blocked — and every decision is written to a SHA-256 hash chain no agent can rewrite.
We are not asking anyone for permission to govern. We are publishing the proof — the code, the audit chain, the doctrine that binds us first — and we are inviting every open-source developer and small AI team running MCP to be a Founding Contributor to the Collective.
The table we are building is open source — and the seats are free. Join us.