Dashboard

Available toAll plans

There is no start or stop button on this page. Locke runs protection on its own, and the Dashboard reports what it observes. The one switch in the app is Pause protection in the title bar, covered in The window.

Protection status

The card at the top left. Its badge reads Running when Locke's detection engine answers a health check, Needs attention when it doesn't, and Checking while the app is still finding out. Under it, one row per part of Locke:

Two lines sit under the rows. Configured to monitor N AI products is how many AI products are switched on in Controls — a setting, not proof that any traffic went through. And while the engine is answering, the card says so plainly: “The engine is responding. This health check does not confirm that your traffic is being screened.” The check talks to the engine directly. It doesn't travel the path your traffic takes, which is why the Desktop Protection and Browser Protection rows are there.

When the detection models need you — not downloaded yet, downloading, a failed download, or waiting for a restart — a Detection models card appears in this column with the button for it (Download detection models, Cancel or Try again). The rest of the time it stays out of the way; the same card lives permanently in Settings › Protection.

While protection is paused

Pausing doesn't add a “Paused” badge to the Dashboard. On Windows, Desktop Protection reads Up · not capturing for as long as the pause lasts, because pausing turns Locke's proxy setting off. On Linux the row keeps reading Up; the title bar's Resume protection button is the sign. Browser Protection stays Connected either way: a pause stops desktop capture only, and the browser extension keeps screening.

Recent activity

The five most recent things Locke did on this device, newest first. Each row reads like email in OpenAI / ChatGPT: the kind of value, the AI service it was headed to, an outcome, and the time. Only labels are listed — never the value itself or anything else you typed. View reports opens Reports; the full live list is on Activity.

Besides detections, a row can mark the first time Locke saw an AI service this session, Locke confirming your surface settings, or a deletion of history on this device. Changes you make to data types or the Never flag list don't appear here — nothing on the device records when those changed.

On a fresh install the card reads No traffic screened yet, with Try the sandbox and See protected apps. That is not a zero: an empty list doesn't confirm traffic was screened and nothing was found. If it reads Recent activity can't be established instead, the app can't read Locke's capture feed right now, and Detailed status and diagnostics says why.

Risk & compliance posture

A one-card summary of the comprehensive compliance checklist that Reports exports, run over this session's outbound detections on this device. Each row counts the checklist's seven requirements, written as N of 7:

Open compliance report goes to Reports for the full checklist. If nothing outbound has been detected this session, the rows read as indeterminate — the card says that is not a finding against you. It is a tool-generated record, not legal advice.

Next step and Shortcuts

The column on the right. Next step suggests the single most useful thing to do, checked in this order:

Shortcuts underneath is always the same four links: Open controls, View reports, Open sandbox and Report an issue, which goes to Support.

The four figures

Four cards, each badged This device. They measure four different things, so don't expect them to add up:

A line under the cards says the same thing the app does: the four aren't slices of one another, only the second is stored history, and all four are your own figures for this device. Locke doesn't show organization-wide totals anywhere.

A dash instead of a number means the app couldn't get a reading — the line under it says why. It is never a zero in disguise.

Detailed status and diagnostics

The row at the bottom of the page, collapsed by default. Click it to open the deeper view. In order, it holds:

Getting started

The first-run checklist, headed Welcome to Locke. Check the detection engine ticks itself when the engine answers; Review configuration ticks once you've opened Controls; and See it protect something real ticks the first time Locke masks or blocks something in a real AI app — a Sandbox test doesn't count for that one. Got it hides the card. To bring it back, use Replay the welcome guide on the Support page.

The engine card and Monitoring status

A note first reminds you that your Controls choices do reach Locke — switching a data type off means Locke still finds the value and sends it in the clear on purpose. Then the engine card: Detection engine is answering, Detection engine is not answering, or Can't determine protection status, with What this check confirms spelling out the health check's blind spot, and Review configuration and Report an issue buttons.

Monitoring status beside it is about the stored history behind the figures: Metrics connected (counts are up to date), Metrics stale (reconnecting, showing the last known counts), Loading metrics, or Metrics unavailable. Under them, a line names your counting settings — for example Counting You typed · occurrences — with a change what counts link to Settings.

Live monitoring feed

The last four detections with their time and outcome, and a tail line — live tail · processed locally, never leaves your device. When nothing new has arrived for a while it says tail is quiet and how long ago the last event was: usually just a quiet spell, but the app can't tell that apart from screening that has stopped, so it tells you the age rather than claiming “live”. An empty feed reads No detection events have reached this feed yet, and — like Recent activity — an empty feed is not a zero.

Sensitive data origins

Sensitive data origins (all-time) breaks your recorded detections down by where each value sat in the request: You typed, Injected context (notes, memory and project files your AI tool attaches), System prompt, Tool output and The AI's replies. Each row shows its share and count, and a pill — in your tally or not counted — so the breakdown always shows the whole picture even when your figures are narrower. Locke protects every row the same way regardless. See How it protects you for why so much of it isn't what you typed.

On a Teams plan

Available toTeams MemberTeams Admin

Everything above works the same on a team: the figures, the feed and the history are your own device's. A few banners are added at the top.

Members see a green notice that the organization configuration is read-only here. Whatever the rest of its wording suggests, your team's policy is applied and enforced on your device — see How a team policy applies. It decides what's protected, and your sidebar has a read-only Team policy page in place of Controls showing what it enforces. Members have no Pause control. When your admin asks you for a report, a second banner names the report and its dates — “You generate it here, on this device; the report itself is never sent to your admin, only that you made it.” Generate report opens Reports set to that period and regime. See For team members.

Admins see a banner reading Manage your team's members, policies, reports and clear requests with Open My Team. The rest of an admin's Dashboard is the admin's own device — members' figures never appear on it. See My Team.