Actra
Agentic Governance
Guardrails
OPA Alternative
Runtime
Claude Skill
Docs
GitHub
Test AI Risk
☀️ Light Mode
Actra Playground
Powered by real Actra WASM compiler in your browser • no backend
✨ Don’t want to write policies manually?
Use
Claude
with the
Actra Claude Skill
to generate policies instantly from natural language.
Download Claude Skill
Schema YAML
version: 1 actions: refund: fields: amount: number actor: fields: role: string snapshot: fields: fraud_flag: boolean
Policy YAML
version: 1 rules: - id: block_large_refund scope: action: refund when: subject: domain: action field: amount operator: greater_than value: literal: 1000 effect: block
Governance YAML
Payload
# Input to policy evaluation actor: role: intern # who is performing action action: type: refund amount: 5000 # what is being attempted snapshot: # optional contextual data about environment/resource resource_owner: external
Decision
Explain
Copy
Validate & Evaluate
How evaluation works
Matched rule
- rule that determined final decision
Failed rule
- evaluated but did not match
Matched condition
- condition evaluated to true
Failed condition
- condition evaluated to false