Docs

Your first FlyDocs workflow: capture to close

Walk through a complete FlyDocs development cycle: start a session, activate an issue, implement, review, and close.

Here is what a full development cycle looks like with FlyDocs. You type slash commands. The agent handles everything else -- loading context, updating your PM tool, formatting code, enforcing process. By the end, you have shipped a feature and your Linear or Jira board is fully up to date without you touching it once.

This walkthrough assumes you have signed up and connected Linear or Jira, installed the CLI, and your workspace is configured. Everything below runs through the cloud tier end-to-end -- every action syncs with your PM tool in real time.

Start a session

AI session 
/start-session

Your agent loads active issues from Linear or Jira, checks the previous session summary, and suggests where to start. You did not explain the project, paste context, or search for the right issue. It already knows.

If you are working in a multi-repo setup, the agent also loads cross-repo context via service descriptors. It knows your service boundaries, API surfaces, and dependencies before you write a line of code.

Capture an issue

AI session 
/capture

Describe what needs to happen. The agent creates a structured issue with acceptance criteria, assigns a type and priority, and pushes it to your PM tool. The issue appears in Linear or Jira immediately -- no browser, no manual entry.

Activate the issue

AI session 
/activate

The agent pulls the issue, transitions it to "In Progress," and assigns it to you -- all in one step. In Linear, the issue moves to the In Progress column. In Jira, the status updates to the mapped equivalent. The agent now has the full spec loaded as context: acceptance criteria, priority, estimates, and any previous comments.

Implement

AI session 
/implement

You focus on the problem. In the background, the agent handles the rest: auto-formatting fires on every edit, context injection keeps the project state current, and script approval runs without confirmation dialogs. As acceptance criteria are met, the agent checks them off in the issue description and posts progress comments -- all visible in your PM tool.

Review

AI session 
/review

Structured review against the acceptance criteria. The agent checks what was built against what was specified, flags gaps, and verifies that your conventions were followed. Not a generic code review -- a spec-aware one. The issue transitions to "In Review" in your PM tool.

Validate (optional)

AI session 
/validate

QE verification from a testing perspective -- edge cases, regressions, acceptance criteria sign-off. The issue moves to "QA" in your PM tool. Skip for smaller changes where review is sufficient.

Close

AI session 
/close

The issue moves to "Done" in your PM tool, outcomes are recorded in the context graph, and the work is archived. In Linear, the issue is marked Done. In Jira, it transitions to the completed status. You did not open either tool. It happened as you worked.

Wrap the session

AI session 
/wrap-session

The agent summarizes what happened, captures decisions, and posts a project update. Next time you run /start-session, this summary loads automatically. Continuity is built in.

Your PM tool reflects exactly what happened. No manual updates. No chasing for status.

Enforcement, not advice

FlyDocs does not suggest standards -- it enforces them. During the workflow above, hooks fire automatically at every step. You do not configure them, invoke them, or think about them. Three layers are working:

  • The stop gate hook prevents the AI from finishing work without handling issue state. It cannot simply complete a task and move on; the issue must be transitioned.
  • Transition validation catches missing comments. Every status change requires a comment explaining what happened and why. Silent status moves are blocked.
  • Auto-formatting runs on every file edit. Code that does not match your project's formatting rules is corrected before the AI continues.

These hooks are managed artifacts -- delivered from the server and kept in sync automatically. You do not install, configure, or maintain them. They are the difference between a framework that structures AI behavior and a set of instructions the AI can choose to ignore.

Next

Join the Discord for support, feature requests, and product updates.