Local execution engine
How the bundled execution engine runs checks and how to diagnose connection issues
The local execution engine runs tests and stores results. It starts with the directly distributed desktop app; normal app use does not require installing a separate server.
Responsibilities
- Manage the test queue and browser processes
- Run configured checks and collect logs, screenshots, traces, and other evidence
- Store reports and provide report queries and analytics
- Collect pages and maintain service map data
- Process interval schedules and permitted MCP requests
The app UI communicates with the engine over a loopback address on the same computer. This describes the internal app connection. Target sites and AI providers can still require external network connections.
App lifetime and automation
Fully quitting the app also stops its bundled engine. Closing a window is not necessarily the same as quitting. An engine that is not running cannot process tests or schedules. See Scheduled checks for sleep and missed intervals.
The desktop engine port is assigned at startup. Do not assume it uses the development server's default port. The MCP bridge discovers the running engine without requiring a hardcoded address.
Diagnose the right connection
| Symptom | Check first |
|---|---|
| App cannot connect to its engine | Engine status and Agent Manager in settings |
| Browser cannot reach the target page | Environment URL, local server, network, and service authentication |
| Only AI tasks fail | Model provider authentication, usage limits, and connectivity |
| Schedule does not run | Engine availability, schedule enabled state, and next run time |
If the connection does not recover, check for running work and restart through Agent Manager, or fully quit and reopen the app. An interrupted run is not a completed verification.
A ready engine does not imply that service login or model authentication is ready. Check these connections separately.