Design partners
Build the governance layer for your domain with us.
CHP is an early, open protocol. The core — declared capabilities, tamper-evident evidence, replay, and structured denials — is real and usable today. The production trust layer for regulated work is what we build next, with the partners who feel the pain first.
We would rather show you exactly how it would work in your domain than claim it already does. That is what this program is for.
Provable today
- Tamper-evident, SHA256-chained evidence for every action — local, no backend.
- Replay any session by correlation ID across parent and child agents.
- Structured denials: an action blocked by policy is a first-class outcome, not an exception.
- Zero-friction capture for AI agents: chp hooks install, no application code changes.
What we build with design partners
- Hosted evidence retention and multi-host trace stitching.
- Role-based access and team workspaces over the evidence store.
- Domain invariant libraries (clinical sign-off, privilege, claims adjudication).
- Compliance and audit export shaped to your regulator and your workflow.
Where we are looking for partners
Pick the workflow you would prove with us.
Each of these is a demonstration of how the protocol would work in a domain — grounded in primitives that exist today. The honest gap in each is the invitation.
AI-native software
Demonstrable todayMake every agent action provable and replayable — and unblock the security review that is holding your launch.
This one already works. Bring a real agent workflow and we will instrument it with you.
Insurance
Design partner wantedTurn automated approve / deny decisions into a provable record of why — defensible in a dispute and to a regulator.
Denial records map almost 1:1 onto the protocol. We would build retention and audit export around your claims flow.
Legal
Design partner wantedA defensible chain of custody for AI-assisted review, with privilege decisions captured as governed approve / deny events.
Tamper-evident chaining is the protocol primitive. We would build matter-level retention and access with you.
Healthcare operations
Design partner wantedAI acts on records; a clinician signs off; both land in one replayable trace with the approval attached.
We would build the access controls and retention an audit actually requires, around a real clinical workflow.
Manufacturing & industrial
Design partner wantedDispatch and machine commands — issued by humans and agents alike — governed by approval and safety invariants, with edge hosts stitched into one trace.
Real-time control stays out of scope. We would prove the governance and evidence layer above it with you.
Financial services
Design partner wantedApprovals, model-risk checks, and AI-in-the-workflow captured as an evidence bundle a reviewer can replay.
We would build the compliance export and controls your risk function needs, on top of the provable core.
Don't see your domain?
If your work needs a provable record of what people, agents, and systems did, the protocol probably fits. Tell us the workflow and we will show you how it would map.