Use cases
What people use CHP for.
Concrete jobs the protocol does. One is provable today; the rest are demonstrations of how the same primitives would work in your domain — each an invitation to build it with us.
See what an AI agent did
LiveThe boundary that is real today. One command hooks your agent CLI and records every tool call as replayable, tamper-evident evidence — the answer to a security review.
See it →Prove why an automated decision happened
DemonstratedDenial is a first-class outcome in CHP, so an automated approve/deny carries an explicit reason, the deciding subject, and an evidence id you can replay.
How it would work →Put a human approval in the record
DemonstratedA sign-off, consent, or authorization becomes the same kind of governed, provable event as an agent action — not a side note in a log.
How it would work →Replay a process across hosts
DemonstratedOne correlation id flows across machines, services, and partners, so a process spanning many hosts replays as a single ordered trace.
How it would work →Gate a high-risk action
DemonstratedA capability can require entitlement, approval, or a safety check; when one fails, the action is denied at the boundary and the denial is recorded.
How it would work →Expose a product capability safely
DemonstratedPublish what your product can do as a governed, discoverable capability with a stable contract and per-invocation evidence.
How it would work →Have a job that needs a provable record?
If your work needs to show what people, agents, and systems did, the protocol probably fits. Tell us the workflow and we will map it.