Concepts
Trust Levels
How Agentronics moves from detection to a verified agent acting for a known user.
Trust Levels
| Trust | Meaning | Reached by |
|---|---|---|
detected | Signals indicate an agent; nothing is proven. | WebMCP activity, vision behaviour, DOM fingerprints, crawler UA |
declared | The agent named itself, or presented a credential nothing has checked yet. | Self-declaration, X-Agent header, unverified tokens in the browser |
verified | A credential was checked. | Web Bot Auth, OAuth 2.1, agent API keys, verified crawlers, or your verifyToken |
linked | A verified agent acting for a signed-in user. | A delegated OAuth token, or a verified session link |
When several methods produce an identity, the SDK keeps the highest-trust one. Use minTrust in authorization rules to decide what each level may call — e.g. verified to place orders, linked to touch a user's account.