Advanced Features

Approvals & Decision Gates

Learn how Burros.AI keeps you in control with approval gates, trust policies, and Decision Gate cards for high-impact agent operations.

Burros.AI is built on the principle of Human-in-the-Loop. While agents are autonomous, they are not unsupervised. Decision Gates are pause points where the system requires your explicit sign-off before an agent can proceed with a high-impact action.

What triggers a Decision Gate?

Common triggers for approval requests include:

  • Destructive Actions: Deleting or overwriting files in the repository.
  • External Network Calls: Requesting data from external APIs or servers.
  • Code Promotions: Moving a deliverable from "Staged" to "Promoted" in the Vault.
  • Critical Commands: Running shell commands that could impact the host environment.

Handling Approvals

When an agent hits a Decision Gate:

  1. The Mission Pauses: Only the specific task requiring approval stops. Other unrelated tasks continue.
  2. Interactive Card: A card appears in the chat showing the command or action requested.
  3. Review & Act: You can Approve (allows the agent to continue) or Reject (stops the action).
  4. Adaptive Feedback: If you reject an action, you should provide a message in the chat explaining why. The agent will read your feedback and try to find an alternative solution.
Tip: Approvals ensure safety without sacrificing velocity. You can review and act on Decision Gates directly from your mobile device or desktop in real time.