App policies

App policies decide how the Endpoint Agent detects and acts on sensitive data in each AI application. They live on the Endpoint Agent tab of the Policy Engine. For how the Policy Engine works across surfaces, see Policy Engine.

Go toQuilrAI consolePolicy EngineEndpoint Agent

The Endpoint Agent tab runs in Basic policies mode: you configure each application on its own card. The tab has two sub-tabs, Detection configurations and Application configuration.

Detection configurations

This sub-tab shows how the agent detects and acts on sensitive data in each application, with a count of active configurations. It lists the supported endpoint applications, for example ChatGPT, Claude, Cursor, GitHub Copilot, Microsoft Copilot, Gemini, DeepSeek, Grok, Slack, Ollama, and LM Studio.

Each application card shows:

Card area	What it covers
Types	The interaction types configured for the app, such as Chat, File upload, Chat anonymous, and Plugin.
Monitoring	Whether the app is monitored on the desktop, in the browser, or both, and the desktop platforms covered (macOS, Windows).
Guardrails	The action per sensitive-data category, for example Monitor, Justify, mandatory justification, or Block.
File restrictions	Rules for files shared with the app.
Group & user rules	Which smart groups and users the configuration applies to.
Monitored browsers	The browsers in which the app's web version is monitored.
Action counts	How many rules justify, require mandatory justification, or monitor.

Application configuration

Use this sub-tab to set what applications are allowed to do on endpoints. It works with process mapping, which enforces execution policies (allow, block, quarantine, justify) on discovered applications.

Example: coach on PII in ChatGPT

1. Open Detection configurations and find the ChatGPT card.
2. Open Guardrails. Set PII to Justify so users are coached and asked for a reason. Use Block for categories such as Auth & Secrets only after you have reviewed the findings, or start everything in Monitor.
3. Open Group & user rules and scope the configuration to a pilot smart group.
4. Check the results in Findings and interactions before widening the scope.

Policy changes reach agents without a restart. See Backend connectivity.

Convert to Policy Engine

The Convert to Policy Engine button moves the Endpoint Agent from Basic policies to the Policy Engine model used by the gateways. Your QuilrAI representative can help you plan the conversion.