# Activity

> The board of everything your agents are doing — what is waiting on you, what is running now, what finished, and how to stop a run without opening it.

**Activity** is the home screen: every conversation across every agent, grouped by what state it is in.

## When to use it

Open it when you have more than one agent, or more than one thing in flight. A conversation tells you about itself; Activity is the only place that answers "what is happening across all of them, and which one needs me first".

It is also where you come back to after being away. [Runs](/docs/work/runs/) continue while the app is closed, so the board is the record of what happened without you.

## The three tabs

The home screen carries three tabs:

| Tab | What it is for |
| --- | --- |
| **Activity** | This board — conversations by state |
| **Usage** | The [usage dashboard](/docs/account/usage/) — tokens, models, agents, over time |
| **What's new** | Product announcements, newest first. A dot on the tab means there are some you have not read |

## Reading the board

<Steps>

1. **Check the counts in the header.** *Active* is how many conversations are running, *Needs review* is how many are blocked on you, *Agents* is how many you have, and the token badge is how much of your plan allowance is spent. The badge links to the [usage dashboard](/docs/account/usage/).

2. **Narrow to the agents you care about.** The row of agent chips under the header filters the board. Selecting none shows everything.

3. **Read the columns.**

   - **Waiting** — the agent stopped and needs you: a [question](/docs/work/questions/), an [approval](/docs/work/approvals/), a skill to grant. Rows here carry a marker, and the column header shows a dot while any of them is blocked on your input.
   - **Active** — a run is in progress. The row names the tool the agent is using at that moment.
   - **Done** — finished, split into **Success** and **Failed**.

4. **Select a row to open the conversation.** That is where you answer, approve, or read the result — the board routes, it does not act.

</Steps>

## Board and Feed

The toggle on the right switches between two readings of the same data:

- **Board** — the three columns above. Use it to triage: what is blocked, what is moving, what is done.
- **Feed** — one list in time order, newest first. Use it to catch up on what happened in sequence.

## Stopping a run from the board

A row in **Active** carries a **Stop** button. It asks for confirmation, then ends that run.

<Aside type="caution" title="Stopping is not undo">
  Work the run already did is kept — messages it sent, files it wrote, actions it
  took. What stops is everything after that point, and a stopped run cannot be
  resumed. To change direction without losing the run, [steer](/docs/work/steering/) it
  instead.
</Aside>

Stop is offered only in the **Active** column. A conversation in **Waiting** is not running — it is holding for your answer — so there is nothing to stop; answer it or leave it.

## Limits

- The board shows your most recent conversations, not your whole history. Older threads live in the [conversation](/docs/work/conversations/) list of the agent that owns them.
- It does not refresh on a timer. Reopening the tab or returning to the page brings it up to date.
- Counts and columns cover conversations, not individual runs. A conversation that ran five times today is one row.
- Filtering by agent changes what you see, not what is counted in the header.

## Related

<CardGrid>
  <LinkCard
    title="Runs"
    href="/docs/work/runs/"
    description="What the Active column is showing you a step of."
  />
  <LinkCard
    title="Notifications"
    href="/docs/work/inbox/"
    description="Being told something finished, instead of checking the board."
  />
  <LinkCard
    title="Usage dashboard"
    href="/docs/account/usage/"
    description="The second tab — where the token badge takes you."
  />
  <LinkCard
    title="Conversations"
    href="/docs/work/conversations/"
    description="Finding a thread the board no longer lists."
  />
</CardGrid>
