How AgentGate works
AgentGate is the machine-readable agent-access policy standard by Ingenious AI. A site publishes one file — /.well-known/agent-policy.json — declaring which AI agents may use it, what they may do, and on what commercial terms. This page documents the format and the tools.
The AGENT-POLICY v0.1 format
robots.txt tells crawlers which paths they may fetch. An agent policy goes further, in four parts:
- Per-agent rules — named agents (and a default rule for everyone else) are allowed, denied, or allowed with conditions.
- Actions — permitted and disallowed actions per agent, such as reading, indexing, training, or transacting.
- Commercial terms — the terms under which agent use is licensed, stated in the file itself.
- Rate limits — request ceilings an agent can read before it calls, not after it is blocked.
The tools
The free generator builds a policy from a short questionnaire — no sign-in, the complete file is the free job. The validator checks any policy file against the v0.1 spec and reports each defect it finds. The conformance page shows worked examples, and the samples are complete policies for common site types. Paid plans add hosted policy URLs and batch validation.
Honest scope
A published policy is a declaration, not enforcement. It states the owner's terms in a form agents and tools can read; it cannot force an agent to comply. Enforcement belongs to the site's own gate (for declared agents, see AgentBlock); AgentGate is the statement those tools and agents read.