Complete a remote task

This example creates a small report on an already paired device and downloads it. It covers target verification, task creation, exit checking and artifact delivery without depending on a particular agent product.

1. Verify prerequisites

Assume endpoint workstation, an existing writable /srv/anytty-demo directory, sh on the target and jq locally. Replace these values:

anytty endpoint show workstation
anytty file stat workstation /srv/anytty-demo --json

Prepare missing directories according to your project and resolve permissions first.

2. Create a named task

endpoint=workstation
task_name="agent-report-$(date +%s)"
created=$(anytty terminal create --endpoint "$endpoint" --json \
  --name "$task_name" --cwd /srv/anytty-demo \
  --tag role=agent,task=report \
  -- sh -lc 'pwd > report.txt; printf "report complete\n"')
target=$(printf '%s' "$created" | jq -er '.target')
printf '%s\n' "$target"

The command writes report.txt; use a dedicated demo directory to avoid overwriting an existing report. Real parallel tasks should use separate output directories.

3. Wait and inspect

anytty terminal wait "$target" --state exited --timeout 2m --json > wait.json
anytty terminal capture "$target" --lines 100 --json > capture.json
jq -e '.exit_code == 0' wait.json

Check each command’s status. If wait fails, stop the success path and inspect whether the task is still running. A nonzero task exit code requires error reporting.

4. Download

anytty file download "$endpoint" /srv/anytty-demo/report.txt ./report.txt --json

Verify transfer success and file content. Real build artifacts may also need project-specific format or content validation.

5. Deliver reviewable evidence

Report endpoint, target, working directory, command, exit code, relevant output and local artifact path. A person can open the same task by name in the TUI manager or App. Keep the record until review is complete, then clean up deliberately.