To see what the agent team is doing, whether it has enough capacity, and what it costs, look in two places: the Agent queue for workload, and Agent management › Usage for usage and cost.
Agent queue
At the top right of Development › Issues, switch to Agent queue to see what the whole agent team is doing. Each agent has its own lane:
| Area | What's in it |
|---|---|
| Now | Work running on a runtime, how long it has been going, and what it did most recently |
| Up next | Queued issues and why they're waiting, such as "No runtime online", "All slots busy", or "Blocked by PUR-12" |
| Waiting for reply | Things waiting on people, such as failed runs, questions, and issues in review, and who they're waiting for. Anything that's your turn is highlighted, and Handle takes you straight there |
| Assigned, idle | Issues assigned to the agent with nothing happening, and why, for example "In Backlog, won't start on its own" |
The top line sums up what's running, what's queued, and what's waiting for you, along with how busy the runtimes are. Check Only mine to show only the issues you own, created, or follow.
If you often see "All slots busy" or long queues, you don't have enough runtimes. Raise a runtime's maximum concurrent runs or connect another one. See Connect a shared runtime. If an agent sits idle all the time, its Good at description may be too narrow for the coordinator agent to think of it. See Enable agents.
Usage
Agent management › Usage lists the tokens agent runs have used and their estimated cost at your model prices:
- Pick a date range, such as the last 7 days, this month, or last month.
- Group by agent, who woke it, subject (issue or session), model, coding tool, or date, or look at just one agent or one person.
- Each row shows the number of runs, duration, input and output tokens, cache reads and writes, and the estimated cost.
- Daily trend charts cost or tokens per day.
Cost per issue on the dashboard uses the same prices. Model calls outside agent runs, such as indexing and searching the knowledge base, are listed separately under Other model use.
Model prices
Costs are estimated from the model prices you set. On the usage page, click Set prices:
- Each coding tool lists the models its runs have reported. Enter input and output prices per million tokens for each model, or price a group of models with a rule.
- For coding tools paid monthly, such as Claude Code signed in with a subscription account, turn on Subscription (not billed), and their runs aren't charged.
- Models without a price show "—" for cost in the reports.
Setting prices needs permission to manage model prices, which usually means an admin.