Policies
Governance Brain
Business law, machine law, approval gates, AI boundaries, and workflow constraints now live in one governed policy experience.
Current Home
The current policy surface lives under Settings > Policies.
Use the policy command center for hard blocks, approvals required, recent evaluations, exceptions, and governed operator fixes.
What Policies Means
Readable operating law instead of hidden rule code.
What is blocked
Why it matters
What evidence is missing
Who can approve
How to fix it
No Commitment Without Quote Token
Customer-facing commitments must be linked to a quote token or validated promise event.
Quotes
Failure code: COMMITMENT_WITHOUT_QUOTE_TOKEN
Active • v1.0.0
No Output Without System Signature
Approved outputs require a signed system envelope for traceability.
System Truth
Failure code: OUTPUT_UNSIGNED
Active • v1.0.0
No Money Without Validation
Refunds, credits, discounts, and settlement actions require finance validation.
Finance
Failure code: MONEY_ACTION_UNVALIDATED
Active • v1.0.0
No Execution Without Ledger Event
External execution requires a ledger event or action intent before side effects begin.
OpenClaw Execution
Failure code: EXECUTION_WITHOUT_LEDGER_EVENT
Active • v1.0.0
No Agent Action Without Command Gate
Agents cannot execute side effects unless the command gate and policy engine have approved the action.
AI Governance
Failure code: AGENT_ACTION_WITHOUT_GATE
Active • v1.0.0
No Exception Without Expiration
Overrides must declare reason, approver, expiration, affected object, and review date.
Policy / Exceptions
Failure code: EXCEPTION_WITHOUT_EXPIRATION
Active • v1.0.0
Capability Envelope Policy
Agents must stay inside their declared domain and action envelope.
AI Governance
Failure code: OUTSIDE_CAPABILITY_ENVELOPE
Active • v1.0.0
Conflicting Data Kill Switch Policy
Critical source conflicts should pause autonomous execution.
AI Governance
Failure code: CONFLICTING_SOURCE_CHANNELS
Active • v1.0.0
Action Intent Required Policy
External execution requires a staged action intent.
OpenClaw Execution
Failure code: ACTION_INTENT_REQUIRED
Active • v1.0.0