Run it yourself. Block it yourself.
Give the coding agent a prompt, then change policies in the GABA console.Verdicts happen on the endpoint within 100ms. Every decision lands in the audit log.
- 1Run the prompt
- 2See the block and the alert
- 3Turn the policy off
- 4Run the same prompt again
Four steps, start to finish
See a block, change a policy, and watch the same prompt get allowed. The tour never clicks for you. It points at a window and tells you what to do.
Overview
Verdicts follow the policies set in the GABA console.
Three things happened at once
Detected by kernel signal, decided locally
The agent's file read is caught as an OS kernel signal. The policy bundle stored on the endpoint decides allow or block within 100ms, and the decision holds even when the policy server is down. On a block, gaba-tray notifies the user.
Policy on behavior, not vendors
Policies attach to behaviors like 'read a sensitive file'. The same action gets the same rule, whatever the AI tool. One toggle pushes it to the endpoint, and an approval workflow handles exceptions.
Every decision is sealed
Blocks and allows alike land in an insert-only log. Who, when, what, which policy matched, and the approver are recorded, meeting the audit-trail requirements of ISMS-P and financial supervisory regulation.
Don't block AI. Make it controllable.
See how GABA Governance works in a 30-minute demo.
