Direct path
An agent calls an operational tool with a reusable credential. A successful API response can be mistaken for a correct business outcome.
Why PALO-AI
PALO-AI extends governance beyond permission. It binds declared authority, policy and human oversight to protected execution, a trusted receipt and verification of the effect that actually occurred.
Developer Preview: use isolated data and reversible, non-consequential tools. This is not a production authorization service or certified security boundary.
The technical difference
An agent calls an operational tool with a reusable credential. A successful API response can be mistaken for a correct business outcome.
The exact claim, authority, approval, capability, receipt and observed outcome become one reviewable chain.
When authoritative post-state contradicts the declared effect, the resource is held and an incident is opened.
Browser-local, synthetic comparison
This interaction explains the control model. It performs no network request, tool call, approval or real execution.
Use arrow keys to move between scenarios.
Evidence state
What PALO changes
Who may act, for what purpose, on which resource and within which limits?
Which exact immutable claim needs review, and who is accountable for it?
Can the action run only through a one-time, claim-bound capability?
What did the trusted executor actually attempt and record?
Does authoritative post-state satisfy the Effect Contract?
If evidence mismatches or is inconclusive, is the resource held for review?
Know the boundary before adoption
Inspect what is implemented, what remains a prototype and the nine independently testable production gates. Then choose the shortest evaluation route for your build mode.