Skip to main content
Long Sessions let an agent work autonomously for a bounded period. They run in an isolated sandbox, stream progress to Dench, save artifacts, ask before risky actions, consume prepaid AI credits, and stop when done or blocked.

​
Start from the CLI

npx -y dench-cli autonomous run "Find useful work to do in this workspace. Do not edit files, publish, deploy, spend money, send external messages, access secrets, or change production data." --duration 30m --json
Use one of these durations: 30m, 1h, 3h, 5h, or until-done.

​
Credits

Long Sessions use prepaid AI credits.
  • If credits are available, Dench starts the run and returns runId, workflowRunId, status, and runUrl.
  • If credits are empty, Dench returns code: "ai_credits_required" and a topUpUrl.
  • If a running session exhausts credits, it pauses. Do not retry in a loop. Ask the human to top up, then resume the same run.

​
Recovery

In --json mode, failed starts return structured errors: 
{
  "ok": false,
  "status": 402,
  "code": "ai_credits_required",
  "topUpUrl": "https://dench.dev/usage",
  "nextActions": [
    "Top up AI credits before starting a Long Session.",
    "Retry the same request after the balance updates."
  ]
}
Follow nextActions once. If the same blocker repeats, report it to the human instead of retrying.

​
Approval rules

Long Sessions still need human approval before risky actions:
  • publishing
  • deploying
  • spending money
  • sending external messages
  • accessing secrets
  • changing production data
If a prompt says not to do a risky action, obey that instruction. Do not request approval just because a risky word appears in a negated safety sentence.

​
Artifacts and suggested work

Use artifacts for durable output: reports, notes, task suggestions, drafts, and patches. Review task suggestions with: 
npx -y dench-cli suggested-work --json
Convert a useful suggestion into a task: 
npx -y dench-cli suggested-work task <artifact_id> --json