Brand intelligence

Agent token

Definition

An agent token is the bearer credential an AI agent presents to read a brand record. It is scoped to named brands, read-only, shown once when it is minted, and revocable the moment you want that agent to stop reading.

Why it matters

Handing an agent access to your brand is a permission decision, not a setup step. The credential decides which brands it reaches, what it may do with them, and how quickly you can take that back.

Most keys get pasted into a config once and forgotten. That holds up until an agent is retired, a contractor moves on, or a secret ends up quoted in a chat window where it will sit for as long as the transcript does.

The safe shape is deliberately boring: read-only by default, narrow in what it reaches, cheap to rotate, and instant to revoke.

How Jinn treats it

A Jinn token is shown exactly once, at mint or rotation, and belongs in a secret manager after that. Minting, rotating, and revoking happen in the web panel rather than over a tool call, so a plaintext secret never lands in an agent transcript.

Rotation keeps the previous token valid for a short overlap window, so agents can be swapped over without downtime, and an explicit revoke takes effect immediately. Multiple named tokens per brand are supported: issue one per agent, and revoking one never disturbs the rest.

Token hygiene, rate limits, and the error codes an agent branches on are documented publicly. See it

Related terms

Browse the full dictionary

A credential you can take back.

Shown once, scoped to your brands, revoked the instant you decide. Read how tokens are minted and rotated.