AI Pods

AI Pods are configurable AI agents that can be assigned to execute backlog items autonomously. Each pod has a defined purpose, a set of tools, and a cost/speed profile.

What an AI Pod Does

When you dispatch a backlog item to an AI Pod, the pod:

1. Reads the item’s acceptance criteria, description, and context
2. Plans an execution approach
3. Executes the plan using available tools (code generation, API calls, search, etc.)
4. Creates an execution run with full logs and artefacts
5. Returns a result — code, document, analysis output, etc.

Pods work best on well-defined, ready items (quality score 80+).

Creating an AI Pod

1. Click AI Pods in the sidebar.
2. Click New Pod.
3. Configure:
   • Name — an identifier for this pod configuration
   • Purpose — the type of work this pod handles (code, analysis, documentation, etc.)
   • Model — the underlying AI model (depending on your subscription and connected provider)
   • System prompt — additional instructions that shape the pod’s behaviour

Dispatching Work to a Pod

From a backlog item detail page:

1. Click Dispatch to Pod.
2. Select the pod.
3. Choose a mode: Autopilot (fully autonomous; agent executes and opens a PR without intervention) or Supervised (agent pauses at checkpoints for human approval before continuing).
4. Confirm.

The pod starts an execution run immediately.

Execution Runs

Each dispatch creates an execution run tracked under Zone 3 (Execution). Runs have full log output, cost tracking (AI tokens + Fargate compute), and phase reporting (PLANNING → EXECUTING → COMMITTING → PR_RAISED → COMPLETED).

For supervised runs, the run pauses at checkpoints with Awaiting approval status — open the run and click Approve to continue or Cancel to abandon.

Navigate to Runs in the sidebar to see all execution history across all pods and projects. See the Execution Runs article for the full reference.

Pod Access Control

Only Admin role users can create or delete pods. All members can dispatch items to existing pods.

Note: AI Pods is an advanced feature. Some capabilities may require a Professional or Enterprise subscription plan.