Skip to content
Index730 resources
Agent settings

Agent settings

The configuration surface that bounds what an autonomous agent may do on the user's behalf.

AI & Agents · 9 checks · 8 resources

Before you ship

9
  1. 01Tool inventory with per-tool enable/disable — every capability the agent can invoke, listed explicitly
  2. 02Autonomy level per tool — always ask, ask once per session, or always allow, with a safe default of always ask for anything destructive or outbound
  3. 03Data scope — which workspaces, folders, repositories or connectors the agent may read, and which it may write
  4. 04Spend and step caps — a maximum token or credit budget and a maximum step count per run, with the behaviour on exhaustion stated
  5. 05Instruction or persona field — persistent custom instructions, with a visible character limit and a preview of the effective prompt
  6. 06Model and reasoning-effort selection — with the cost and latency implication of each option shown
  7. 07Schedule and trigger configuration — when the agent runs unattended, and what event starts it
  8. 08Kill switch — a single control that halts all running agents and revokes their tokens immediately
  9. 09Audit trail link — where to review what the agent actually did under these settings

Resources that help you build this

8