Skip to content
Documentation
Open the panel

Command lineCommands

What an agent does

The overview, sessions, transcripts, costs, usage, memory, logs and events of an agent, and the dashboard of every agent. The panel shows the same on the Overview, Sessions, Costs, Memory and Live tabs of an agent and on the Agents tab of each project.

1

Step 1: See what an agent does and how it goes

The overview tells what the agent works on or did last, what needs attention, how its work of the last 7 days went and what it cost, as the Overview tab of the agent in the panel does. Tasks with the same source label, which the monitor gave them when it queued them, stand in one group, and a task without a source stands alone. --json prints the whole answer.

$ kcloud agents overview sandbox/release-notes

Output

NOW
Last: #142 (issue-142) Fix the flaky login test, done, finished 2026-03-14T09:30:00.000Z
Result: The login test waits for the session cookie instead of a fixed delay.
Queue: 1 pending, 1 failed
Last review of new work: 2026-03-14T09:30:00.000Z, found nothing new

NEEDS ATTENTION (3)
succeeded on attempt 2: ENG-102 (eng-102-impl) Send the shipping notifications by SMS
handed off to a person: PROJ-88 (proj-88) Accept returns without a receipt: A decision on refunds without a receipt is needed.
failed after 3 attempts: ci-9182634 Failed test run 9182634: The session ran past its time limit.

RECENT WORK
SOURCE        TITLE                                   TASKS  OUTCOME             TIME   COST  FINISHED
#142          Fix the flaky login test                    1  done               7 min  $0.84  2026-03-14T09:30:00.000Z
ENG-102       Send the shipping notifications by SMS      2  done              22 min  $2.33  2026-03-14T09:30:00.000Z
eng-102-impl  Send the shipping notifications by SMS         done, 2 attempts  17 min  $1.92  2026-03-14T09:30:00.000Z
eng-102-plan  Plan the shipping notifications by SMS         done               4 min  $0.41  2026-03-14T09:30:00.000Z
PROJ-88       Accept returns without a receipt            1  handedOff          3 min  $0.36  2026-03-14T09:30:00.000Z
ci-9182634    Failed test run 9182634                     1  failed            30 min  $0.97  2026-03-14T09:30:00.000Z

HOW IT GOES
WINDOW    DONE  FAILED  HANDED OFF  TASK COST  AVERAGE      REVIEWS
24 hours     3       1           1      $4.50   11 min   48 ($1.44)
7 days      12       4           4     $18.00   11 min  192 ($5.76)
Cost today: $3.17
Last session: monitor, 2026-03-14T09:30:00.000Z
2

Step 2: See what the agents are doing

The dashboard shows each agent of the projects you can see: its status, since when, its current task and the cost of the current UTC day. With --follow the command prints a line for each change after the table, until Ctrl-C.

$ kcloud agents dashboard

Output

PROJECT  AGENT          STATUS   SINCE                     TASK  COST TODAY
sandbox  nightly        stopped  2026-03-14T09:30:00.000Z  -           0.00
sandbox  release-notes  waiting  2026-03-14T09:30:00.000Z  -           0.00
3

Step 3: List the sessions

The command shows the Claude sessions of the agent, the newest first, with their role, model, outcome, failure and cost. --role and --task narrow the list.

$ kcloud agents sessions sandbox/release-notes --limit 5

Output

STARTED                   ROLE      MODEL             OUTCOME  FAILURE  COST  SESSION
2026-03-14T09:30:00.000Z  executor  claude-opus-5-5   ok       -        0.84  0195a3c4-7e21-7b6d-9f40-3c81a4d2e600
2026-03-14T09:30:00.000Z  monitor   claude-haiku-5-5  ok       -        0.03  0195a3c4-7e21-7b6d-9f40-3c81a4d2e601
2026-03-14T09:30:00.000Z  monitor   claude-haiku-5-5  ok       -        0.03  0195a3c4-7e21-7b6d-9f40-3c81a4d2e602
2026-03-14T09:30:00.000Z  executor  claude-opus-5-5   ok       -        0.84  0195a3c4-7e21-7b6d-9f40-3c81a4d2e603
2026-03-14T09:30:00.000Z  monitor   claude-haiku-5-5  ok       -        0.03  0195a3c4-7e21-7b6d-9f40-3c81a4d2e604
4

Step 4: Read the transcript of a session

Name the session by its id from the SESSION column of the list of sessions. The command prints what the agent wrote and which tools it used.

$ kcloud agents transcript sandbox/release-notes 0195a3c4-7e21-7b6d-9f40-3c81a4d2e600

Output

{"kind":"init","ts":"2026-03-14T09:30:00.000Z","model":"claude-opus-5-5","tools":24,"mcpServers":[{"name":"playwright","status":"connected"}]}
{"kind":"text","ts":"2026-03-14T09:30:00.000Z","text":"I will run the login test a few times to see how it fails."}
{"kind":"toolUse","ts":"2026-03-14T09:30:00.000Z","id":"toolu_01","name":"Bash","input":{"command":"pnpm test --repeat 10 test/login.test.ts"},"truncated":false}
{"kind":"toolResult","ts":"2026-03-14T09:30:00.000Z","toolUseId":"toolu_01","isError":false,"text":"9 passed, 1 failed: expected the session cookie before the redirect","truncated":false}
{"kind":"text","ts":"2026-03-14T09:30:00.000Z","text":"The test reads the cookie before the redirect finishes. I will wait for the redirect first."}
{"kind":"toolUse","ts":"2026-03-14T09:30:00.000Z","id":"toolu_02","name":"Edit","input":{"file_path":"test/login.test.ts"},"truncated":false}
{"kind":"toolResult","ts":"2026-03-14T09:30:00.000Z","toolUseId":"toolu_02","isError":false,"text":"The file test/login.test.ts has been updated.","truncated":false}
{"kind":"result","ts":"2026-03-14T09:30:00.000Z","ok":true,"costUsd":0.84,"durationMs":412000,"numTurns":23,"text":"The login test passes 50 runs out of 50. I opened a pull request with the fix."}
5

Step 5: Check the costs of an agent

The command counts the costs day by day over the last --days days, and task by task with --by-task.

$ kcloud agents costs sandbox/release-notes --days 7

Output

The agent spent nothing between 2026-03-14 and 2026-03-14.

0.00 over 0 runs and 23 sessions.
6

Step 6: Check what the container used

The command shows the CPU and memory of the container of the agent over --range: 1h, 24h or 7d.

$ kcloud agents metrics sandbox/release-notes --range 1h

Output

AT                          CPU   MEMORY
2026-03-14T09:30:00.000Z  18.0%   12 MiB
2026-03-14T09:30:00.000Z  22.6%  138 MiB
2026-03-14T09:30:00.000Z  30.5%  174 MiB
7

Step 7: Search the memory of an agent

The agent remembers summaries of its tasks. --query searches for a text, and --outcome keeps the tasks that succeeded (succeeded) or failed (failed).

$ kcloud agents memory sandbox/release-notes --limit 5

Output

The agent remembers nothing that matches.
8

Step 8: Read the logs

The command prints the last lines of what the agent writes and leaves when the output stops. --limit sets how many lines to replay first.

$ kcloud agents logs sandbox/release-notes --limit 10

Output

2026-03-14T09:30:00.000Z info  - worker.started {"ts":"2026-03-14T09:30:00.000Z","event":"worker.started","level":"info","version":"0.12.0","pid":7,"nodeVersion":"v22.22.0","paused":false,"authKind":"oauth","projectDir":"/workspace"}
2026-03-14T09:30:00.000Z info  monitor control.changed {"ts":"2026-03-14T09:30:00.000Z","agent":"monitor","event":"control.changed","level":"info","state":"running"}
2026-03-14T09:30:00.000Z info  executor control.changed {"ts":"2026-03-14T09:30:00.000Z","agent":"executor","event":"control.changed","level":"info","state":"running"}
2026-03-14T09:30:00.000Z info  monitor cycle.started {"ts":"2026-03-14T09:30:00.000Z","agent":"monitor","event":"cycle.started","level":"info","cycle":1}
2026-03-14T09:30:00.000Z info  monitor admission.granted {"ts":"2026-03-14T09:30:00.000Z","agent":"monitor","event":"admission.granted","level":"info","ticket":"monitor-1","seq":1}
2026-03-14T09:30:00.000Z info  monitor cycle.finished {"ts":"2026-03-14T09:30:00.000Z","agent":"monitor","event":"cycle.finished","level":"info","cycle":1,"ok":true,"durationMs":38000,"costUsd":0.03,"addedTasks":1,"skippedDuplicates":0}
2026-03-14T09:30:00.000Z info  monitor admission.released {"ts":"2026-03-14T09:30:00.000Z","agent":"monitor","event":"admission.released","level":"info","ticket":"monitor-1","seq":2}
2026-03-14T09:30:00.000Z info  executor task.started {"ts":"2026-03-14T09:30:00.000Z","agent":"executor","event":"task.started","level":"info","taskId":"issue-142","title":"Fix the flaky login test"}
2026-03-14T09:30:00.000Z info  executor task.finished {"ts":"2026-03-14T09:30:00.000Z","agent":"executor","event":"task.finished","level":"info","taskId":"issue-142","title":"Fix the flaky login test","ok":true,"durationMs":412000,"costUsd":0.84,"numTurns":23}
2026-03-14T09:30:00.000Z info  executor executor.waiting {"ts":"2026-03-14T09:30:00.000Z","agent":"executor","event":"executor.waiting","level":"info"}

With --follow (-f) the logs stream live until Ctrl-C stops them, as on the Live tab of the panel. kcloud agents logs sandbox/release-notes --before <event> reads earlier lines, and the command prints at the end how to read the page before.

9

Step 9: Read the events

Events are the recorded steps of the agent: monitor cycles, sessions, tasks and changes of state. With --follow they stream live, and --since <event> resumes reading where you stopped.

$ kcloud agents events sandbox/release-notes --limit 10

Output

2026-03-14T09:30:00.000Z info  - worker.started {"ts":"2026-03-14T09:30:00.000Z","event":"worker.started","level":"info","version":"0.12.0","pid":7,"nodeVersion":"v22.22.0","paused":false,"authKind":"oauth","projectDir":"/workspace"}
2026-03-14T09:30:00.000Z info  monitor control.changed {"ts":"2026-03-14T09:30:00.000Z","agent":"monitor","event":"control.changed","level":"info","state":"running"}
2026-03-14T09:30:00.000Z info  executor control.changed {"ts":"2026-03-14T09:30:00.000Z","agent":"executor","event":"control.changed","level":"info","state":"running"}
2026-03-14T09:30:00.000Z info  monitor cycle.started {"ts":"2026-03-14T09:30:00.000Z","agent":"monitor","event":"cycle.started","level":"info","cycle":1}
2026-03-14T09:30:00.000Z info  monitor admission.granted {"ts":"2026-03-14T09:30:00.000Z","agent":"monitor","event":"admission.granted","level":"info","ticket":"monitor-1","seq":1}
2026-03-14T09:30:00.000Z info  monitor cycle.finished {"ts":"2026-03-14T09:30:00.000Z","agent":"monitor","event":"cycle.finished","level":"info","cycle":1,"ok":true,"durationMs":38000,"costUsd":0.03,"addedTasks":1,"skippedDuplicates":0}
2026-03-14T09:30:00.000Z info  monitor admission.released {"ts":"2026-03-14T09:30:00.000Z","agent":"monitor","event":"admission.released","level":"info","ticket":"monitor-1","seq":2}
2026-03-14T09:30:00.000Z info  executor task.started {"ts":"2026-03-14T09:30:00.000Z","agent":"executor","event":"task.started","level":"info","taskId":"issue-142","title":"Fix the flaky login test"}
2026-03-14T09:30:00.000Z info  executor task.finished {"ts":"2026-03-14T09:30:00.000Z","agent":"executor","event":"task.finished","level":"info","taskId":"issue-142","title":"Fix the flaky login test","ok":true,"durationMs":412000,"costUsd":0.84,"numTurns":23}
2026-03-14T09:30:00.000Z info  executor executor.waiting {"ts":"2026-03-14T09:30:00.000Z","agent":"executor","event":"executor.waiting","level":"info"}

Matches the panel at cloud.kikimora.dev.