Playbooks
Playbooks turn detected opportunities into steps such as detect, execute, review, ship, and measure. Each run activates one intelligent-action candidate. Your organization must have Playbooks enabled. Every command requires--scope <org-slug>/<product-slug> (a dashboard URL also
works). Authenticate with tpc auth login first. Output defaults to a text table;
pass --format json for API data.
Catalog and settings
list shows every playbook’s ID, name, category, availability (built or coming soon), enabled, auto-run, and Autopilot settings, and active run count. Missing settings show
unset; this does not imply a server default. Counts cover only the latest 50
runs, not all historical runs.
set <playbook-id> accepts --enabled=true|false, --auto-run=true|false, and
--autopilot=true|false. Provide at least one. Omitted fields stay unchanged;
explicit false is sent to the API. Enabling auto-run allows the server to start
runs automatically when eligible candidates appear. Autopilot publishes waiting
drafts without review, up to 3 pages a day per playbook; drafts over that limit,
and every draft while the product has no connected domain, wait for your review.
Coming-soon playbooks cannot execute yet.
Find candidates and start a run
suggestions reads the intelligent-actions bootstrap and lists available
candidates whose action type maps to a playbook. The text table includes the
candidate ID, playbook ID, action type, and title. An empty table means no matching
candidate was returned; detection may still be materializing.
run start executes the selected candidate. Use the playbook and candidate IDs
from suggestions. The server validates eligibility and rejects duplicate live
runs for the same activation with HTTP 409. Inspect existing runs before retrying.
Inspect runs
runs returns the overview’s latest 50 runs, newest first. --playbook <id> and
--status <status> filter that window locally; they do not fetch older runs.
Statuses are running, needs_review, shipping, measuring, recovered,
not_recovered, completed, failed, and cancelled. The first four are active.
run get shows the status, steps, measurement checkpoints, and errors. At review,
it links to the existing draft review page when the run resource identifies a
draft; otherwise it links to the product’s review queue. Approve or reject there.
Playbooks has no separate CLI approval command. The engine observes draft review
and publication state.
Measured playbooks capture a baseline at start and checkpoints at day 7 and day 14
after shipping. They end recovered or not_recovered. Other runs can end
completed, failed, or cancelled.
Cancel a run
JSON output
data without adding display fields:
list: the overview object withsettingsandruns. The static catalog is in the text table; it is not added to the API response.runs: the filtered run array, preserving all run fields.suggestions: the filtered candidate array, preserving all candidate fields.run get,run start,run cancel: the run object, including steps and measurements.set: the saved setting object.
resource.type = "agentic_document_draft" and resource.id identifies the draft at
/<org-slug>/p/<product-slug>/docs/review/<draft-id>.
Errors
HTTP 404 can mean Playbooks is not enabled for the organization or the scoped resource does not exist. Check scope and IDs withlist / runs, and ask your
organization administrator about access. HTTP 409 means the request conflicts
with current state; inspect runs before retrying. Authentication and permission
errors retain the server’s explanation.