Metrics, integrations, privacy, personal workspace, and CLI.
Try the or write to us on Telegram
Neither. The system provides telemetry similar to services like Google Analytics, AppMetrics, and others. It does not take screenshots, record the screen, or read messages. The goal is to supply data about work processes and help find and remove bottlenecks.
Instead of reports and subjective assessment, an employee can show the blockers they hit while working on tasks and back up their position.
A manager, in turn, can rely on data when making management decisions — from finding process bottlenecks to defending their decisions to senior leadership.
How to talk to the team before installing plugins and reduce resistance: how to get employee support when introducing telemetry.
Code, secrets, keys, and passwords are not read or transmitted. We collect activity metadata:
The IDE is the primary signal, then it is joined with Git and the tracker.
Time in development is focused work against the company’s daily coding norm. The default is 1 hour per day; a user with the Maintainer or Owner role can change it on the chart.
Coding, CLI and database time count. The system looks at the gap between neighbouring signals: 15 minutes or less — the silence between them is still work; more than 15 minutes — a break, and those minutes are not hours. Browser and other categories do not appear on the chart.
The same rule applies to humans and AI: short agent tasks with gaps longer than 15 minutes between them do not stitch together.
Density shows how much a person’s events overlap with AI-agent events. It includes coding, database, CLI, browsing, and agent events.
Occupancy on this widget is the union of all that time. It is the same number as on the “Activity time” widget, without meetings: calls are added there separately. The overlay cap in the UI is ×2, meaning the agent ran alongside the person for the whole occupancy. The value is computed from period totals.
By signal source: the event came from manual coding or from an AI agent. The plugin identifies the source and tags it. If someone uses an AI tool without a PanDev plugin, the system counts it as manual coding, which inflates that figure. This compromise keeps the user’s data from being dropped.
There is no manual correction or exclusion of such data. For AI activity to be attributed correctly, you need a plugin for the tool.
This metric is the share of “branch + file” pairs with AI coding that also appear in the merge request file history. The source is file activity on the branch plus Git change history.
Zero can show up when there are no AI-touched files in the period, or none of them made it into branch history. Spikes appear where that overlap showed up.
A correct count needs a Git integration, activity with a file name and AI-coding events, and those same files in the branch change history.
The “Time in development” metric is meant as focused work on one task in code: you can already see where someone stays on a task and where they bounce.
If tracking breaks after a tool update, a patch usually ships within a week. A new integration, once prioritized, takes about two weeks. Plugins cover the tool’s version line and update automatically. The integration list is in the matching PanDev section.
With PanDev Tracker we are starting to cover every application companies use, through a single app.
Right now an administrator can set up MCP queries to the system to check, on a schedule, for interrupted signals from individual employees. There are no in-product notifications.
Zoom is already supported — step-by-step setup is in PanDev integrations. Google Meet goes through the browser plugin.
We support Zoom, Google Meet, and Microsoft Teams. Any integration we do not have yet we can add in about two weeks.
Activity on a git branch is linked to an issue by key. The key is matched with regexes in the branch name, MR title, and commit message. If the key is missing from those places, matching will not work.
The basic rule for the system to work correctly is following git flow
There is no after-hours collection cutoff. After the corporate network drops, signals also buffer on the device and flush on the next connection.
There is no “corporate laptop only” bind. Our recommendation for keeping work and personal apart is a work profile in each app, or splitting OS accounts into work and personal.
What else the system regulates:
- browser extensions have a blacklist of resources whose visits will not be sent to telemetry even from a work account
- in browser extensions you can see the organisation whitelist. How it works: when an event appears, the extension first checks the domain against the whitelist, then the blacklist. If both lists contain the domain, the send is rejected.
- git repositories that are not work repos are not sent
After the connection to the corporate network is lost, telemetry accumulates on the device and will be sent when the connection is restored
The employer sees exactly what the employee sees in their own profile. Telemetry is sent only from apps and sites the developer allowed.
A manager sees telemetry only for their own team. Team members see only their own data. There is no separate consent to disclose data to colleagues; access is governed by role.
Roles: Owner, Maintainer, Viewer, Member, Team Lead, Finance. You cannot create custom roles, and you cannot change the permission set inside a role.
Access is defined by the role; available actions:
If you are already signed in, open Account settings from the profile menu and click Change password: you will need the old password and a new one. If you cannot sign in — Forgot password? on the login screen, the email you registered with, and the link from the letter.
With AD/LDAP the password is changed in the corporate directory; contact the team that administers it.
There is no dedicated overtime metric in the UI. Period AI analysis can already flag systematic overtime.
Regular reports with those signals can be assembled through PanDev MCP using an external agent.
Personalizing the set and order of widgets on dashboards is not supported. Custom dashboards can be built with Datavisor or via MCP
There are no separate personal goals, badges, or team challenges right now. Profile gamification is in the SaaS version on the personal plan for small teams: levels, stats, and an achievement showcase.
There is no public roadmap. We usually publish a date for an integration or fix in the support chat after an estimate.
You can request a feature or report a bug in the support chat, by email, or with the “Report a bug” button in the bottom-right corner of the product UI.
We collect feedback, improve specific scenarios, and publish usage scenarios on a dedicated page
Spend is attributed to specific issues from the branch name: feature/PAY-1204 → PAY-1204. The branch used is the one you were on at the time of the work, so it is critical to follow git flow
If you work outside a branch, the activity is treated as unattributed. If you worked on a branch with an arbitrary name, main, dev — that is, without a key — the system will show how much work happened on that branch, but will not attach an issue or DORA metrics to it.
Unattributed spend is split into “in a repo, branch has no key” (rename the branch and it attributes itself) and “outside a repository” (there is nothing to attach to). The share of such branches is visible in the UI
The CLI plugin supports Claude Code, Codex CLI, Cursor, Windsurf, Google Antigravity, ZCode, opencode, and Kilo Code.
With the PanDev Tracker desktop app, any tool used on the work machine is supported
PanDev does not send any requests outside the virtual machine — no logs, telemetry, or personal data. The only access the system needs is to our licensing service. That is required to check that the subscription is current and valid.
Also, if an error occurs on the client, a user of the system UI can send it to us with the matching button right from the UI.