> ## Documentation Index > Fetch the complete documentation index at: https://docs.cogniagent.ai/llms.txt > Use this file to discover all available pages before exploring further. # Watching progress > See exactly what your AI coworker is doing as it works — every step, its checklist, and what it produced. Nothing is hidden. You can watch every step as it happens, or ignore it entirely and read the recap. Both are fine — but the option matters, because it's how you learn to trust the thing. ## What you see As your coworker works, the conversation fills with a row per step. Searched for something. Opened a file. Ran the numbers. Sent a message. Each row is a summary you can expand for the detail — what it actually searched for, what came back, what it changed. Nothing is hidden behind a spinner. You'll also see rows where it was **thinking** rather than doing. Those are collapsed by default; expand one when you want to know why it made a particular choice. A sequence of step rows showing a search, a page read, a calculation and a file being written ## The checklist On a job with several parts, your coworker keeps a visible to-do list and ticks items off as it goes. This is the fastest way to tell whether a long task is going well. If the list matches what you expected, you can stop watching. If item two is something you didn't ask for, you've caught it early — say so and it adjusts. The plan checklist with three completed items, one in progress and one pending ## Special cards Some steps get their own card rather than a plain row, because they're worth seeing properly. | Card | What it is | | ---------------------- | ------------------------------------------------------------------------------------------------------------- | | **Approval** | Something needs your yes. See [Approvals](/cowork/control/approvals) | | **A question for you** | It's stuck on something only you know | | **Browser handoff** | It hit a login and wants the keyboard. See [The live browser](/cowork/capabilities/live-browser) | | **Delegation** | It handed work to another coworker. See [Working with coworkers](/cowork/capabilities/working-with-coworkers) | | **Workflow run** | A team of helpers is working. See [Dynamic workflows](/cowork/capabilities/dynamic-workflows) | | **File change** | It edited a file, and you can see exactly what changed | | **Image** | It generated a picture, shown inline | ## Finishing When it's done, your coworker posts a quiet wrap-up line you can expand for the full account of what happened. It's deliberately understated. The work is in the conversation above and in the files panel; the wrap-up just tells you it's finished and what the outcome was. ## When you were away If your coworker worked while you weren't watching — because you closed the tab, or because it [scheduled its own follow-up](/cowork/capabilities/self-scheduling) — you get a notice saying it woke up, why, and what it did. Nothing happens off the record. Everything it did while you were away is in the conversation, in order. ## Good to know **Watch the first few tasks closely, then stop.** The point of watching isn't supervision, it's calibration — learning where this coworker is reliable and where it needs a tighter brief. Every coworker also has an **Activity** tab with its complete history across all tasks, if you want the wider view rather than one conversation. ## Common questions No. Close the tab; it keeps working. The conversation is all there when you come back. Yes. Stop the task and it halts. Everything done so far stays, including any files it produced. Usually a hard problem, or a high [effort setting](/cowork/chat/effort-and-model). Expand the thinking row to see what it's weighing. Yes, line by line. See [Seeing what changed](/cowork/files/changes). Type a correction. It's picked up as soon as the current step settles, without stopping the task. If it's badly off, stop it and re-ask. ## Next steps The cards that need your yes. Several jobs at once. Checking its work. When a task runs long.