Every document, published in full

What the software can record

Not all of this is necessarily switched on for you. Your employer chooses the configuration, and it can turn most of it off. Ask your employer for its own notice, which should tell you what is enabled where you work.

• —Which applications and websites are open, and the titles of the active window or page.
• —When you are active and when you are idle, and how long working sessions last.
• —How intensely you are using the keyboard and mouse — a count of key presses and clicks, turned into an activity level.
• —Screenshots of your screen, taken at intervals, if your employer has enabled them. They can be blurred, and specific applications and sites can be excluded before anything leaves your device.
• —Live frames of your screen while a manager has a live view session open, if your employer has enabled live view.
• —If your employer has turned on the AI tool telemetry relay: usage telemetry from AI coding tools you run on the computer, such as Claude Code or Codex. The tools send it through the WorkMonitor agent, which passes it on and ties each session to you as the person the computer is enrolled to. It covers when a session ran and for how long, which AI models it used, how many tokens it spent and what it cost, the repository and branch it worked in where the tool is set to report them, and a record of each step of the session: what kind of step it was (a call to an AI model, a tool the AI used, a permission decision or an error), when it happened, how long it took, and which model a model call used. For a session you drive yourself, that record keeps which tool was used only when it is one of the AI tool's own built-in tools, such as running a command or editing a file, and an error's standard code where the tool reports one, but it never keeps the name of any other tool. It keeps how each step ended only where the tool reports it in one of a few fixed words (ok, error, cancelled, denied or timeout) or as a standard status code, so the record can show which call failed or was denied; any other description of how a step ended is dropped. The session also keeps how many errors it had, failed tool calls included. For an AI tool set up to work on its own, such as one started from a script, it keeps all of these in full, with the name of any tool, how each step ended in the tool's own words and whether each step was allowed, and your employer can set an AI tool you drive to keep the same full record. It does not include the text of your prompts, the replies or your code: WorkMonitor's settings never ask a tool to send them, and WorkMonitor does not record them. Where your employer has set rules for its AI tools, WorkMonitor also checks each step it records against them afterwards, whether the step came through the agent or straight from the tool: a step that matches is flagged on that session, with the rule, when it happened, the kind of step, the repository, and the tool's name where it is one of the AI tool's own built-in tools or the rule names it, never the model, and a rule can stop that session, or every session you drive with that AI tool.
• —If your employer has turned on AI tool discovery: which AI coding tools from a fixed list run on the computer, and whether each is set up to report to WorkMonitor. To tell, the agent looks at the programs running under your account and reads the reporting settings in those tools' own settings files. Only the result leaves the computer: the tool's name and how it reports, never a list of your programs and never anything else in those files.
• —If your employer has also chosen to configure AI tools on each computer: the agent adds WorkMonitor's reporting settings to those AI tools' settings files on your computer, so that they report through it. It changes only its own entries and leaves a setting you chose yourself alone. It removes its entries when your employer turns this off, when you sign out, or when the computer is removed from your employer's account. If the agent is uninstalled without signing out first, its entries stay in those files until they are removed.
• —None of these three AI coding tool settings runs unless you have given your consent.
• —Separately, your employer can set AI coding tools or AI agents to report the same usage telemetry, with its record of each step, straight to WorkMonitor without the WorkMonitor agent: for example through a company-wide setting on your computer, or from an automated pipeline. This path does not ask for your consent in WorkMonitor. When such a report carries an email address or WorkMonitor member identifier that matches your WorkMonitor account, the session is attributed to you.
• —If your employer has turned on AI policy checks, the check looks at what an AI coding tool you drive on the computer is about to do before it acts: running a command, reading or changing a file, opening a web address or using a tool of another service your employer connected. It compares the action with your employer's rules, and the action can be stopped, held until you confirm it in the AI tool or until someone your employer allows approves it, or only flagged. The check reads the command, the file path or the web address only to compare it with your employer's rules and WorkMonitor's own safeguards, and keeps none of them. Through the WorkMonitor agent, the command leaves the computer without the values the software can recognise as secrets, such as a password written into it; a secret typed as a separate word may still be sent, and is not kept. Where your employer sets an AI tool to check straight with WorkMonitor without the agent, for example on a server, the tool sends everything it was about to do, including any text it was about to write, and WorkMonitor reads only the session, the tool's name, the command, the file path, the web address and the working folder from it, and keeps none of them apart from the session and the tool's name as described below. To know which repository the AI tool is working in, the check reads that repository's own Git settings on the computer.
• —Each time an action is stopped, held for your confirmation or an approval, or flagged, WorkMonitor records when it happened, which AI tool it was, the kind of action, the name of the tool the AI used where it is one of the AI tool's own built-in tools or the rule names it, the rule, stop or safeguard that applied, the repository, your session and the computer, and never the command, the file or the web address. For an approval, it also records who answered it and how (approved, denied or expired), and any reason they gave. Nobody can approve a request made through the WorkMonitor agent for their own session. For an approval a check asks for, the people asked see the rule, the AI tool, the repository and, in WorkMonitor, whose session it is, but never the command, and WorkMonitor never adds your name to a Slack message about it; an AI agent that asks for approval itself shows them its own description of what it wants to do. If you answer approvals for others, your answers and reasons are recorded too.
• —Your employer, or one of its rules, can stop an AI coding session you drive, or every session you drive with that AI tool, and WorkMonitor records the stop, why it was made and when it was lifted. Where your employer selects it for a stop, the stop can also remove your seat at the company that provides the AI tool, and WorkMonitor keeps the details of that account.
• —AI policy checks do not ask for your consent in WorkMonitor. They keep running while capture is paused and during private time, so your employer's rules and stops still apply then, but a check made through the WorkMonitor agent then is not recorded with you, your session or your computer. That includes your calls to tools of other services your employer connected through WorkMonitor, and no one else's approval is asked then. Checks an AI tool makes with WorkMonitor directly, without the agent, are recorded with the session the tool names, because without the agent WorkMonitor cannot tell whose session it is, or that you are paused or in private time. WorkMonitor counts them as monitoring when it checks your employer's setup against the notice and consultation rules where you work. To show your employer whether the checks work on your computer, the agent reports how each AI coding tool there is set up for them. Where your employer has also chosen to configure AI tools on each computer, the agent can add the entries that run these checks to Claude Code's own settings file, beside any of your own, and removes them when your employer turns either setting off or when you sign out.
• —If your employer uses WorkMonitor to check whether its AI agents' work held up: when you review a pull request WorkMonitor keeps, including one kept to track work, or merge one that an AI agent made or helped make, what you did, when, and your GitHub account, linked to you where your employer has confirmed that the account is yours. Managers who can see that AI agent see how many reviews each person made on its pull requests, and they see your name there only if you are among the people they are allowed to see; everyone else is counted together. That count is not part of any score about you. A pull request you opened is also kept, with its title, its link and your GitHub account, when its description, its branch name or its commits carry a mark that an AI tool made or co-wrote it (only the mark is looked for, and the description and commit messages are not kept), when the repository, branch and time of your AI coding sessions match it, or when an AI agent reports it as its work. It is then linked as work an AI agent did or helped with: confirmed where the AI tool's own record names the pull request, and otherwise it can be suggested for a manager to confirm or reject. When an AI agent hands work to you by name, the handoff and its time are recorded.
• —If your employer uses WorkMonitor to manage what its AI tools cost: what the AI tools it pays for cost for you, as seats and as usage, and which paid seats you hold and when each was last used. Managers who can see AI spend see it by name only for the people they are allowed to see. Your employer can set budgets on this cost, including a budget for one person or for each person, and the people it chooses (by default, whoever set the budget) are alerted when a budget nears or passes its limit. When your employer turns on stops, a per-person budget can pause your AI coding sessions. A paid seat you have not used for 30 days, or a seat on a second tool of the same kind that you did not use, can be suggested for cancelling, and using two tools of the same kind can be noted for a manager.
• —For AI coding tools you drive, WorkMonitor keeps their rates of use to catch a session that runs away: tokens, steps, actions and sessions per minute, and cost per hour. These are rates and counts, never the content of a session. When a rate goes far above its usual level, or above a fixed limit, which your employer can change, the person your employer made responsible for that AI tool in WorkMonitor, if there is one, and the people your employer chose are alerted, and a runaway incident is recorded: when it happened, which AI tool, which rate and its peak. Where your employer turns on stops, a runaway incident can also pause the session. An alert about a budget or a session names you only to someone allowed to see you, and WorkMonitor never adds a person's name to a Slack message about a budget. The budget's own name, which your employer chooses, is shown in every budget alert.
• —If your employer records a pay or cost rate for you, or a billing rate for your work, it is shown to people allowed to see your member record or the rate cards, which includes managers by default, to people allowed to see payroll or invoices, to people allowed to see AI agents, in the evidence behind an invoice, to people allowed to administer the organization, in its audit log, which records each rate added to the rate cards, with its amount, and in that log's exports to a security log and through the API, and to people allowed to see performance cases, in a check-in that prices your recovered hours. A billing rate is also shown beside your name on the invoices for your work, to the client and in your employer's accounting system. WorkMonitor uses it for payroll, to price client invoices and the time sent to your employer's accounting system, for cost and margin reports and project budget reports, to price any shortfall against your expected hours in your weekly performance record and what a check-in later recovers of it, to price unusual timesheets before they are approved where your employer turns this on, and, in AI cost, only for the quarterly average cost of a merged pull request across the whole organization, worked out once for each quarter, after it ends, from people's tracked time on those pull requests. Reports that show a priced amount beside a person's time, such as a project's budget against its time by member, findings on unusual timesheets and custom reports on performance weeks, let their readers work out that person's rate. The average is shown only when it comes from at least 20 pull requests and at least three people, with no one person above 35 percent of the time or of the cost. It is never shown for one person, and only people who can see AI spend across the whole organization and can already see those rates see it.
• —Your tracked time on projects is used to share the cost of the AI tools you use among those projects and their clients, unless your employer turns this off. When your employer sends those costs as invoices to its accounting system, the invoice lines name projects, clients and AI tools, never a person.
• —Your working time: clock-ins and clock-outs, timesheets, attendance, and the projects and tasks you record.
• —If your employer connects GitHub to track work: the pull requests and issues the connected GitHub account opened or is assigned to, with their title, link, state and dates and the GitHub accounts of their author and assignees, never their description. They can be matched to the people who worked on them and to their tracked time and activity; a GitHub account is linked to a person only when your employer confirms it is theirs.
• —Where your employer uses geofenced job sites, the GPS position recorded at the moment you clock in or out on a device that supports it.
• —Technical information about the device: its name, its operating system, the version of the agent, and the IP address it connects from. When you sign a computer in, or confirm it, through your browser and the computer offers a key of its own, the network your browser was on is kept with that computer for 90 days, and in the audit record of that sign-in for as long as audit records are kept; it is a network prefix, not your full address.
• —If you sign in with a mobile telephone number rather than a work email address, that number. We send a six-digit code to it when you sign in, and we use it for nothing else — it is not used to locate you, it is not shared with your employer's advertising, and it is not part of anything the software records about your work.

What the software cannot do

These are statements about what the software is capable of, not promises about how it is used. They are the things you can hold us to.

• —It does not record the text of what you type. Keystrokes are counted, not captured. There is no keylogger in the product and no code path that stores the content of a key press.
• —It does not record audio. There is no microphone capture anywhere in the product.
• —It does not use your camera or record video of you.
• —It does not make continuous screen recordings. Where screenshots are enabled they are still images at an interval; live view frames exist only while a session is open and are discarded when it closes.
• —It does not read your email, your messages or your documents. It records that an application was in the foreground and what its window was called, not what the document says. The only files of yours it reads are the settings files of known AI coding tools, while your employer has turned on any of the three AI coding tool settings above or AI policy checks, and only to check or set their reporting settings or to see how the checks are set up, and, for AI policy checks, the Git settings of the repository an AI coding tool is working in.
• —It does not follow you around the internet outside work, and it does not run on devices your employer has not deployed it to.
• —It does not track your continuous location. The only location the platform records is the position at a clock-in or clock-out event, where your employer uses job sites.

• —One honest caveat, because it matters: window and page titles are recorded, and a title can be revealing — the name of a document, the subject of an email, the name of a clinic in a browser tab. That is why the software lets your employer exclude applications and sites, and blur captures, before anything leaves your device. If you think something private is being captured, this is the thing to raise, and it can be fixed by configuration.

Who can see it

Inside your organization: the administrators and managers your employer has authorized, within the boundary it has set. Permissions can be scoped, so a manager may be able to see their own team and no further.

Every time someone opens captured material about you, that access is recorded on a tamper-evident log. The people watching are themselves watched, and your employer can produce that record.

At WorkMonitor: our staff can access customer environments only where it is needed to run or support the service, under confidentiality obligations, and that access is logged and reviewable by your employer.

Outside: the service providers we use to run the platform, listed in full on our Sub-processors page, and any system your employer has connected — an HR or payroll system, for example. Your employer chooses those connections.

We do not sell data about you. We do not use it to train AI models, and the AI providers we use are contractually prohibited from training on it.

Advertising, stated precisely because we do advertise. WorkMonitor buys advertisements on Google, and our website measures how those advertisements perform — which of them bring people to us, and which lead to a sign-up. That measurement uses data about the person browsing — the pages they looked at, and the advertisement they arrived from. It does not use, and has no route to, anything the software records about you: not an application or window title, not an activity figure, not a screenshot, not your hours, not a score, not your name. Nothing captured on your device is used to advertise anything to anyone, and no advertising system ever receives it.

The one nuance worth being straight about. If you sign in to WorkMonitor in a browser and accept cookies there, you are a website visitor for that purpose like anyone else, and the measurement sees which pages that browser opened. That is your choice rather than your employer's, and you can refuse it at any time: the "Do Not Sell or Share My Personal Information" link in the footer turns advertising off on one click and leaves everything else alone. What is captured on your device travels a different path entirely and never joins it.

And one more, if you are also the person who bought the subscription. When an account that came to us from one of our advertisements pays its first invoice, we tell our advertising provider that the purchase happened, with the amount and the advertisement click it came from — and without the buyer's email address in any form, scrambled or not. That is about the purchase, not about you as a monitored person, and it still carries nothing the software recorded on your device. If you are not the buyer it does not concern you at all, and no monitored person's address is ever sent in any form.

Scores, summaries and AI

The platform turns activity into productivity, focus and engagement scores, and can generate written summaries of a day or a week using AI.

These are measurements of signals from a device. They do not know that you were thinking, reading on paper, in a corridor conversation that solved the problem, or having a bad week for a reason that is none of the software's business.

Our rules for employers say that a score or an AI summary must not be the sole basis for a decision about your discipline, dismissal, pay, promotion or performance rating, and that a person with the authority to disagree has to review the underlying data first. If a decision has been made about you by software alone, that is a matter to raise with your employer, and in many countries you have a specific legal right to object to it.

The underlying data is available in the platform, so a summary you disagree with can be checked against what it was drawn from.

What AI tools cost for you, your budgets, your AI tools' rates of use and the quarterly average cost of a merged pull request never feed your productivity, focus or engagement scores.

Controls you may have

Depending on what your employer has enabled, the platform can give you a self-view of what is captured about you, a private-time pause that stops capture for a period, a consent record you can see, and a route to dispute something.

We build those controls. Your employer decides whether to switch them on. If you cannot find them, ask your employer — the capability exists.

How long it is kept

Your employer sets the retention period, within the limit of its plan: up to 30 days on the free plan, up to 180 days on Pro, up to 365 days on Business, and a period it chooses on Enterprise.

A sweep runs every day and deletes captures past that window. That deletion is permanent — it cannot be undone by your employer or by us.

AI coding tool data has its own periods: the record of which AI coding tools a computer runs and how each reports is deleted 90 days after the computer last reported it, and at the computer's next report once AI tool discovery is turned off for you; the step-by-step detail of AI tool sessions is deleted after the period your employer's plan sets; and session totals (when, for how long, which models, tokens, cost and errors, and the repository and branch) are kept for the life of your employer's account.

Records of AI policy checks, approvals and stops are kept for at least six months, or for the period your employer's plan sets for the step-by-step detail of AI tool sessions where that is longer, counted from when a decision was made, an approval was answered or a stop was lifted; an approval still waiting for an answer and a stop still in force are kept until they end. Actions that were only flagged are kept for the period your employer's plan sets for the step-by-step detail of AI tool sessions.

Your employer can place a legal hold on records it needs for a legal claim. A hold names the matter, covers AI agents, repositories, a client, named people or the records about something such as a claim, over a period, and must give the date its period ends or a date by which it is reviewed. While it stands, the records of AI policy checks, approvals, stops and AI tool sessions it covers are kept past their usual retention, until your employer releases it.

Records of whether AI agents' work held up, including the reviews and merges recorded for you, are kept for the life of your employer's account, and so are the pull requests and issues from work tracking.

AI cost data has its own periods. What each AI tool billed, and the paid seats assigned to you, are kept for the life of your employer's account, as financial records. Daily cost lines, the record of each budget's alerts and states, and the in-app alert notices about budgets and runaway sessions are kept 25 months. Runaway incidents are kept 400 days after they close. The spread of per-minute rates behind the usual levels of use is kept 16 days, and the daily totals of use 35 days; the live readings expire within three days. Savings suggestions are deleted 90 days after they are resolved or their dismissal ends. The quarterly average cost of a merged pull request is kept two years after its quarter.

Live view frames are discarded when the session ends and are never added to the stored history.

Your rights, and who to ask

Depending on where you live, you may have the right to be told what is held about you, to get a copy, to have it corrected, to have it deleted, to object to processing, and to complain to a regulator.

Ask your employer first. It is the controller: it holds the decisions, it has the tools in the platform to answer you, and the law generally directs your request to it. Our agreement with your employer obliges us to help it respond to you.

The export your employer makes of your data also includes the records of AI policy checks made while you drove an AI coding tool or on your computers, the approvals asked for your sessions and those you answered, the stops on your sessions and those you made or lifted, any seat a stop removed for you, the legal holds that name you, the connected services you were given access to through WorkMonitor, the computers you signed in or confirmed with a key of their own and how each is set up for the checks, and the AI control changes you made yourself, such as rules and holds. Erasing your data removes you and your computers from those records and lifts the stops on your sessions. Where a legal hold covers some of those records or names you, your links to them, and the hold's mention of you, are kept until the hold is released, and erased then. Meanwhile they are restricted to the legal matter: everyday screens show a person kept under a legal hold instead of your name, and only the people who manage AI agents across your employer's whole organization see those links, through the hold's own export.

The export your employer makes of your data includes the reviews and merges recorded for you, the pull requests and issues your confirmed GitHub account opened or is assigned to, and the work AI agents handed to you by name. Erasing your data removes you, and any GitHub account your employer confirmed as yours, from the reviews and merges recorded at that moment, removes that account from the pull requests and issues it appears on, and removes you as the reviewer of any link between a pull request or issue and someone's work. A link already made between a pull request or issue and your work stays. A later read from GitHub can record that account on a pull request or issue again, but never link that account to you again.

The export also includes the paid seats you hold and when each was last used, your AI cost by month, the costs you entered, budgets about you with their alerts and states, the stops a budget or a runaway incident asked for that name you, the alerts you were listed to receive and those that name you, the runaway incidents and daily rate totals of sessions you drove, and the savings suggestions and cost rules that name you. Erasing your data removes you from all of these: the money stays in your employer's totals without you, and the quarterly average cost of a merged pull request already worked out stays, because it names no one. A paid seat keeps the AI tool's own account id, so the seat count stays right, but it is never linked to you again.

We cannot hand over, correct or delete your employer's data because you asked us to. That is not a brush-off — it is the same rule that stops anyone else asking us for data about you.

If you write to us at privacy@workmonitor.ai anyway, we will acknowledge you, tell you plainly what we can and cannot do, and pass the request to your employer without undue delay unless doing so would be unlawful.

Two things are not limited by any of that. You can complain to your own data protection authority or regulator at any time, without going through your employer or us. And if you believe your employer's use of WorkMonitor is unlawful — monitoring nobody was told about, monitoring outside working hours, monitoring aimed at union activity or a protected complaint — you can tell us at privacy@workmonitor.ai. Our Acceptable Use Policy prohibits all of that, we will look at what you tell us, and we will not identify you to your employer where we are not obliged to and where doing so could expose you to retaliation.

What your employer had to do before monitoring you

We set this out so you know what to ask about. Under our Acceptable Use Policy, before deploying WorkMonitor an employer must have a lawful basis for the monitoring, must have given you the notices the law requires, and must have completed any consultation, works-council agreement, employee-representative approval, regulatory notification or impact assessment that applies where you work.

Covert monitoring is prohibited on our platform except in narrow, lawful, documented circumstances.

Monitoring aimed at trade-union activity, whistleblowing, a health-and-safety complaint, a discrimination complaint, or any other legally protected activity is prohibited outright, as is monitoring designed to identify your religion, health, disability, pregnancy, sexual orientation, gender identity, race or ethnic origin.

Whether your employer actually did these things is something we do not audit and cannot confirm. What we can tell you is that it agreed to, and that we will act on a credible report that it did not.

Contact

Your employer, first, for anything about your own data.

WorkMonitor: privacy@workmonitor.ai, or our Data Protection Officer at dpo@workmonitor.ai. By post: Digital Socket LLC, 131 Continental Dr, Suite 305, Newark, DE 19713, United States.

Our full Privacy Policy explains our role in more detail, and the Data Processing Addendum sets out what we owe your employer.