allow, then the hold and deny variants.
Other paths: Choose your AtlaSent quickstart.
Developer / API
Evaluate a sandbox action, handle a bounded Permit, verify it, and inspect the evidence that actually exists.
Enterprise operator
Take one protected action from organization setup through rehearsal and readiness review.
AI agent / MCP
Put AtlaSent between an agent tool request and a consequential effect without treating tool access as authority.
The lifecycle you are learning
Verified identity → organizational authority → authorization → Permit → verification → execution → evidence Each stage establishes something different. Anallow does not prove execution. A verified Permit does not prove the downstream effect occurred.
Existing developer sequence
The detailed developer references remain available:The Developer / API quickstart uses a demo action that executes nothing. Console keys are live keys, so its requests send
context.environment: "production"; that names the key’s environment and activates nothing. The other paths have their own safety guidance: follow the page you choose. Protecting a real production action is a separate, controlled step.