Run Your First Task

Once a runner is online, Flowy follows a simple loop: create a task, assign it, watch it run, and review the result.

Step 1: Open a project or use Inbox

You can work in either place:

• Projects keeps related work together.
• Inbox shows active tasks across every project.

For a first run, opening a project is usually the clearest path.

Step 2: Create a task

1. Click New Task.
2. Enter a short, direct title.
3. Use the description for context, constraints, or the expected outcome.
4. Set priority and labels if they help with triage.
5. Click Create task.

Good tasks usually include the goal, the repo or workspace involved, and any limits the AI should respect.

Step 3: Assign the task

Open the task, then click Assign task or Reassign.

In the assignment dialog:

1. Choose a runner.
2. Choose an AI provider from the list that runner advertises.
3. Fill in the harness settings you want to pass through.
4. Click Assign task.

If a provider is missing, that runner does not currently detect the matching CLI.

Step 4: Set the right harness values

The most important field is usually workspace.

• Set workspace to a path that exists on the runner machine.
• Set model only if you want to force a specific model.
• Set sandbox, mode, or worktree only when the selected CLI needs them.

You do not need to fill every field. Flowy works well when you keep the config as small as possible.

Step 5: Watch execution

After assignment:

1. The task is queued on the selected runner.
2. The runner picks it up and changes the task to in_progress.
3. Output streams back into Flowy while the CLI runs.
4. The runner status changes to busy.
5. When the command exits, the task becomes done or failed.

The Inbox view is useful when you want to monitor active work across multiple projects at once.

Step 6: Review the result

Open the task detail view to inspect:

• streamed output
• assigned runner
• selected provider
• harness summary badges
• final status

If a task fails, you can update the task details and assign it again to the same runner or a different one.

A simple working pattern

Many teams settle into this sequence:

1. Create a project for a repo or workstream.
2. Add focused tasks with clear titles.
3. Route each task to the machine that has the right CLI and workspace.
4. Use Inbox to watch active work.
5. Move completed tasks to done and retry failures with better context.

For help choosing runners, providers, and assignment fields, continue to Runner & Provider Guide.