Auto mode

Ship a story end to end.

Start collaboratively when the scope needs discussion. Use autonomous mode when the story has clear acceptance criteria. Your coding client executes the work; Storybloq guides the stages and saves progress. In Codex chat, use $story in place of /story in the examples below.

Written guide available · Video coming soon

Follow along

  1. 1.

    Collaborate: think it through, then execute

    Ask your agent to recommend a story and discuss the scope and acceptance criteria. Work collaboratively, or explicitly invoke /story auto T-001 in Claude Code or $story auto T-001 in Codex to enter the autonomous workflow.

  2. 2.

    Targeted auto: /story auto T-001

    Invoke autonomous mode with a well-scoped story ID. The agent plans, requests the configured reviews, implements, tests, and finalizes. Permissions, blockers, or failed checks can still require your input.

    $/story auto T-001
  3. 3.

    Full-backlog auto: /story auto

    Got a whole phase of scaffolding to burn through? Run /story auto with no story ID. The machine picks the next unblocked story, ships it, picks the next, and keeps going until the backlog drains or it hits something that needs you. Set recipeOverrides.maxTicketsPerSession in .story/config.json to limit the run; the default is 0, meaning unlimited.

    $/story auto
  4. 4.

    Inspect the plan and its review

    PLAN_REVIEW uses your configured backends to challenge the approach before implementation. Storybloq saves an immutable snapshot when the reviewed plan advances. Human approval requirements depend on your workflow and instructions; review the plan before letting the agent proceed when you want that checkpoint.

  5. 5.

    Watch it run

    The agent follows the configured test and review stages, then finalizes the work. The Mac app displays recorded progress and session state. Review backends and stage settings determine which checks run; elapsed time varies with the work.

  6. 6.

    Inspect the commit and handover

    FINALIZE can create the commit before the session reaches COMPLETE. Inspect the resulting commit, test and review evidence, and handover. If the run continues, it can select the next eligible story. Do not assume changes remain uncommitted until you return.