Tracking

A system console for the editor. See what the app is doing, check that each part of it is healthy, and try a capability out — all on your own device.

Open it from the Tracking icon in the activity bar, with ⇧⌘Y / Ctrl+Shift+Y, or from View ▸ Trace Viewer… on the desktop app.

What you can do

Trace — watch activity as it happens. Every event shows its name, source and time; the one you pick opens as neatly formatted JSON beside the list. Filter by text, by type (info, result, error, progress, audit), or follow a single chain of related events from end to end. Export a copy of the log to a file.

Health — one card per part of the system: extensions and the capabilities they register, connected MCP servers and their tools, compute providers, the current policy mode and its approvals, and a breakdown of recent activity. Anything this build of the app cannot report says so, rather than showing a misleading zero.

Capabilities — pick any registered capability, give it some input, run it, and jump straight to the trace of what it did. Runs go through the normal path, so the usual permission rules still apply.

Your data stays local

Nothing is sent anywhere — the console only reads activity already on your device, and makes no network calls. Common secrets such as API keys, tokens and passwords are masked in the detail view and in exports, though this masking is best-effort and cannot catch everything, so treat an exported log as sensitive.

Good to know

The editor keeps only the most recent 200 events in memory, so the Trace list is a recent window rather than a full history. Clear empties that log for good.

Getting started

Run System Console from the command palette.