My work

My work is your personal queue. Open it from the sidebar to see everything routed to you in one place: agents waiting for a human to unblock them, sessions asking for network access, advance requests that need your approval, active tasks you created, tasks you’re watching, and the third-party apps and machines connected to your account.

My work has seven sections, in this order: Blockers, Egress approvals, Awaiting your approval, Created by me, Watching, Connected apps, and Your machines. They don’t all share the same scope: Blockers and Created by me cover the workspace you have active plus any workspaces nested beneath it; Egress approvals covers the current workspace only; Watching spans every workspace you belong to, not just the active one; Connected apps and Your machines are account-level — they don’t scope to a workspace at all, since they list what’s connected to you, not to any one workspace. My work differs from two other workspace views:

• The Dashboard shows a whole-workspace pulse: live sessions, stalled tasks, workspace-wide activity, and cost and throughput charts. Use it for an at-a-glance summary of everything happening across the team.
• The Tasks view is the complete workspace task list — every task, regardless of who created or owns it. Use it for browsing, filtering, and triaging work across the team.

Blockers

An agent raises a blocker when it cannot continue without a decision only you can make. The Blockers section lists every open blocker that needs a person. Notes about unmet dependencies are left out, because they clear themselves. Each entry shows the affected task, the reason the agent stopped, who raised the blocker, and when.

To clear a blocker and let the agent resume, click Resolve next to the entry. You can also open the task first to read the full context — and if the agent is still running, the live terminal is available there. For the full blocker lifecycle — every way a task gets a blocker, which ones flip it to blocked, which clear themselves, and how to file and resolve one from the CLI — see Blockers and blocked tasks. For the agent-question route to blocked specifically, see When an agent needs your input.

Egress approvals

When a session asks to reach a network destination it isn’t allowed to use yet, the request waits here. The Egress approvals section lists pending network access requests for the current workspace only, not workspaces nested beneath it. Requests for the same destination are grouped together.

Only a workspace admin or owner can decide a request. Everyone else sees the buttons disabled, with the reason. For the three answers and what each one does, see Network access approvals.

Awaiting your approval

When a task’s workflow has an approval gate, CodeHerder’s own automatic advance is what files a pending-advance request — not a command a person or agent runs. The Awaiting your approval section lists every pending request where you are an eligible approver, showing the task title, the stage the task would move to, who requested it, and when.

Click the task row to open the task detail page, then use the Approve or Reject buttons in the pending-advance banner. For the full approval workflow, including CLI commands to approve or reject, see Approvals & staying in control.

Created by me

The Created by me section shows every task you filed that isn’t finished yet, at any stage, including one you just filed. Click any row to open that task.

To see finished tasks you filed too, add --all from the CLI:

ch task list --created-by me --all

Watching

The Watching section lists tasks you’ve subscribed to that aren’t finished yet, across every workspace you belong to. The watch mode you chose decides which changes send you a direct message (DM). A watch set to none sends no DMs, while other modes add more events, such as comments.

Click any row to open the task. To subscribe, unsubscribe, or change which events DM you, see Watching tasks and notifications.

From the CLI, add --all to see finished watched tasks too — without it, only unfinished ones are listed:

ch task list --watching --all

Connected apps

The Connected apps section lists every third-party client you’ve approved to act on your CodeHerder account through the Model Context Protocol (MCP) — for example, a claude.ai or ChatGPT connector you’ve connected. It shows the app’s name, its access (Read only or Read & write), and when you connected it. If you haven’t connected anything yet, this section is empty; approving a connector from claude.ai or chatgpt.com lists it here the moment you approve it.

Open the More menu on an app’s row and choose Revoke, then confirm, to cut off its access. That client can no longer act on your behalf — if you (or whoever set it up) want to reconnect it later, you’ll need to approve it again from claude.ai or chatgpt.com. This section is account-level: it shows every connected app across your whole account, not just the current workspace.

Your machines

The Your machines section lists every machine or key currently signed into your account. A row comes from one of two sources, shown in the Source column:

• CLI login — a machine where you’ve signed in with ch login. It shows the machine’s hostname, its operating system, and when the CLI last made a call from it. A machine that’s never made a call yet shows no “last used” value.
• API key — a plain API key, for example one an admin minted for you with ch member keys create and handed over directly. It shows the key’s label (its --name, blank if the key has none) and its last-used time. Its OS cell is always empty, which is the quickest way to tell a key row from a machine.

A key that matches a ch login installation’s current credential is not listed twice. If you haven’t signed in from anywhere and hold no API key, this section is empty. Signing in with ch login, or being handed an API key, lists it here.

Open the More menu on a row and choose Revoke, then confirm, to cut off its access. Revocation and recovery differ by source:

• A CLI login row revokes the installation. Run ch login again from that machine to sign back in — the row reappears.
• An API key row revokes the key itself, permanently. Its plaintext was shown once, at mint time, and a revoked key never comes back by signing in again. To keep working, someone with Admin or owner access at your home workspace must mint you a fresh key with ch member keys create — see Minting API keys.

Use Revoke if you lose a laptop, retire a machine, or don’t recognize an entry. Like Connected apps, this section is account-level, not scoped to a workspace. ch logout only handles the machine you’re currently on — see Credentials and profiles — revoking a different machine or key is done here, in the web app.

The sidebar badge

The My work link in the sidebar shows a count badge for tasks that need your attention right now: every task you created or are watching that isn’t finished yet, de-duplicated so a task you both created and watch counts once.

The badge counts every stage your workspace’s types treat as unfinished, plus blocked; it leaves out only tasks that have finished. The badge is hidden at zero, and shows 99+ once the count passes 99.

From the CLI, these sections have an equivalent command:

ch egress list --status pending     # pending network access requests
ch task list --awaiting-approval     # pending advances awaiting your approval
ch task list --watching --all        # every task you're watching, finished ones included
ch task list --created-by me --all   # everything you've filed, at any stage

---

For approvals and gate configuration, see Approvals & staying in control. For responding to an agent that needs your input, see When an agent needs your input. For subscribing to tasks and managing notifications, see Watching tasks and notifications. For network access requests, see Network access approvals. For the full task list and filter options, see Finding and tracking your work. For what a connected app can do and what ch login sets up, see What Claude and ChatGPT can do with CodeHerder, Connect CodeHerder to claude.ai or ChatGPT, and Credentials and profiles. For minting a replacement API key, see Minting API keys.