Who we are
An independent referee only counts if you can check it yourself.
MakerChecker is the independent referee and evidence standard for AI agents taking high-consequence actions. A company cannot issue its own financial audit, and an agent cannot self-certify. Our self-hosted, domain-neutral software sits outside the model and refuses, in code, to let the agent that proposed an irreversible action approve its own work, a hard 403, not a prompt-level promise. Every action commits to a signed, hash-chained log your own auditor recomputes offline, with no access to your systems. It fits any regulated domain, from banks to life sciences teams to anywhere a step cannot be undone.
Why we build it in the open.
The whole point of MakerChecker is that it is the outside party. The record stands on its own math, not on our reputation: your own auditor recomputes the Ed25519 signatures and SHA-256 hash chain offline, with no access to your systems. So the code is open source and self-hosted, your data never reaches us, and your security team can read every line before you run it. We would rather be checked than believed.
We work with a small number of design partners at a time, close to the people who carry the accountability, so the product is shaped by the real job and not by a roadmap written in a vacuum.
Want to talk to us directly? Email hello@makerchecker.ai, or read how we handle trust and security.
See it for yourself
See an agent get stopped.
One command starts the demo: an agent stopped from signing off its own work, and the signed evidence file an inspector can check for themselves.
Designed against the rules your auditors already enforce.