← All practicals

Register an AI agent, then kill it

Give an AI agent its own identity — scoped, secreted, and instantly revocable.

1. 1

   On the Agents lab, register a new agent with a name and some scopes.

   An AI agent is a non-human identity (NHI): its own row, its own scopes, its own secret — shown to you ONCE and stored only as a SHA-256 hash, exactly like an API key.

2. 2

   Test its access — it comes back active.

   The test widget authenticates the secret through the SAME function every later agent feature will use: hash it, look it up, check status === 'active'.

3. 3

   Hit the kill switch.

   One call flips the agent to 'killed' — an atomic, owner-scoped flip, not a soft flag some routes might forget to check.

4. 4

   Test its access again — it is now dead.

   The exact same authentication call now returns nothing. There is no other code path an agent secret can succeed through, so a kill here is a kill everywhere, instantly.