Hey Product Hunt 👋

I built Prove Before Act after noticing a recurring problem with autonomous agents :

You can have detailed logs of what an agent did and still have no independent proof that its declared decision context existed before the action.

Logs can tell you what happened. But if the system controls the logs, they don't independently prove what was committed before execution. That's the gap PBA is designed to close.

The pattern is simple:

WHY → PROOF → ACT → WHAT → PROOF

Before execution, the agent commits a cryptographic proof of its declared WHY.

After execution, it can anchor evidence of WHAT actually happened.

The underlying data can remain private while the commitment and verification evidence remain independently verifiable.

This isn't just a spec. xproof_agent_verify has been running live on Moltbook for months, using xProof to anchor agent decisions on-chain.

What started as an experiment in making blockchain proofs accessible to agents gradually became something larger :

a pattern for verifiable agent commitments.

As autonomous systems gain the ability to move money, deploy code, call APIs and coordinate with other agents, accountability becomes increasingly important.

Runtime safety and policy enforcement answer one question :

What was the agent allowed to do?

PBA addresses another :

What did the agent commit to before it acted, and can we independently verify that afterward?

PBA is an open standard. xProof is the reference implementation.

You can try it through MCP with 10 proofs free, with no wallet or credit card required.

I'm especially interested in feedback from people building autonomous agents, agent infrastructure or multi-agent systems. And I have one question I'd genuinely love builders here to answer :

What action would you refuse to let an agent take without evidence established beforehand ?

Jason
Solo maker, Prove Before Act