A gateway and ledger for autonomous AI agents
It holds credentials, writes to databases, calls external tools. When something goes wrong, the only record is a chat transcript. In April 2026, an agent deleted a production database and “confessed.” The top reply was: your agents are your responsibility.
Before the agent touches the world, its action is submitted to the NOMOS gateway. The verdict is allow, block, or escalate to a human, evaluated against real framework controls and data-governance regulations, not vague best practices.
Every decision lands in a hash-chained, tamper-evident ledger. Not a chat log, not a dashboard screenshot, runtime evidence of what the agent actually did, structured against the same frameworks an auditor will ask about.
Before an autonomous agent touches the world, its action is submitted to the NOMOS gateway. The verdict is allow, block, or escalate to a human, evaluated against recognized framework controls and data-governance regulations, not vague best practices.
Allow, block, or escalate, each decision lands in a hash-chained ledger, tamper-evident by construction and ready to hand to an auditor. Runtime evidence of what an agent actually did.
Agents act fast, act autonomously, and fail differently from conventional software. Guardrails govern what a model says. An AI model is only as secure as the system around it. When you give AI access to your tools, credentials, and multi-step processes, you create new security risks. NOMOS exists to solve this exact challenge.
Every action waits for approval. Safe, but it doesn’t scale, and the queue becomes the bottleneck.
Autonomous agents act without per-step approval, fast, but nobody can say afterward exactly what they did.
Governed autonomy: agents act freely inside the checkpoint, and every decision is evidence. The aspiration, an agent that governs agents.
The intellectual thesis behind NOMOS is published and peer-reviewed. These papers established the transparency-first security analysis and the privacy-enhancing technology framework that the gateway and ledger are built on.
A transparency-first analysis of AI security through the confidentiality-integrity-availability lens. The thesis: transparency evidence is itself a security control. This is the intellectual lineage of the NOMOS ledger, not a paper about NOMOS, but the framework it descended from.
Read via DOI →Where large language models in ICS/OT cybersecurity create a privacy paradox, and which PETs close the gap for critical-infrastructure operators. The governance challenge NOMOS addresses: agents operating in regulated environments where data residency and access controls are non-negotiable.
Read on IEEE Xplore →Built and verified: the gateway, the hash-chained ledger, trust scoring and circuit breakers, with the control crosswalk across NIST AI RMF, ISO/IEC 42001, the EU AI Act and HIPAA.
Done & verifiedThe checkout, license-key delivery and download path that puts the system in your environment, the last layer before general availability.
In preparationFirst installs run with direct support from the team building NOMOS, and help shape what ships.
You’re hereNOMOS is a local Docker download. It runs where your data lives. Goose FL never hosts it and never touches your data.
Policies and evidence map to recognized frameworks and regulations.
Every policy decision is recorded in an immutable ledger, and the evidence packages it produces are what you bring to an ISO 42001-style audit or an AIUC-1 assessment.
Talk to us about putting NOMOS to work for you as soon as it launches.