Guardian Agent
This card lives in Policy Engine > LLM Gateway at web.quilr.ai/policy. Edits join the shared draft and take effect once you review and publish a revision.
Turn on Guardian's coding helpers and task-adherence checks for matching requests. Runs at the request stage.

Sections
Applies on assistants, bedrock, chat, copilot, responses, sdk_check and vertex.

Settings
Every add button opens the Guardian settings dialog. Each control is a three-way switch, so one configuration can change a single setting and leave the rest to a broader one.

Leave as is keeps the value from a broader configuration. At least one control must be On or Off.
Follow-up rules
Add follow-up opens Guardian follow-up. When is one of: Guardian blocked, at least N findings (N from 1), or a finding type. The rule then tags a severity. Guardian results are request-stage fields; they are not visible on the response leg. The highest severity among matching rules is reported.
Example
runs on requestpriority 700
Scoping and precedence
- Applies to: Everyone, People, Smart group, Application, App tag, Requested model, Provider, API surface, Environment, Prompt text, Tool, Source network, or Except....
- When several configurations match, the highest-priority one wins per setting.
- Off on a narrower scope beats On for Everyone, so you can exempt one app or group.
- Need Guardian only for certain languages, repos or request metadata? Use Add configuration at the bottom of the card.
Legacy app setting
Replaces each app's Guardian Agent settings section while the Policy Engine is on.