Skip to content
Foundry

Orchestration

What it is

Section titled “What it is”

Orchestration coordinates multiple AI agents working in parallel on a set of tasks. You define the scope (sprint, workstream, or custom task list), and the orchestration engine generates a wave-based plan that assigns agents to tasks with dependency awareness. Each run tracks progress in real time with status badges, activity feeds, and a dependency graph.

Why it matters

Section titled “Why it matters”

Dispatching agents one-by-one does not scale. Orchestration automates the coordination problem — it figures out which tasks can run in parallel, assigns the right agent to each, and monitors the entire fleet through a single view.

Key concepts

Section titled “Key concepts”
  • Orchestration wizard: OrchestrationWizard guides run creation through three steps:
    1. WizardStepScope — select sprint, workstream, or custom tasks plus target repositories
    2. WizardStepAssignments — review and adjust AI-generated agent-to-task assignments
    3. WizardStepPreview — preview the full plan before launching
  • Wizard progress states: creating_run, generating_plan, starting — shown with elapsed timer
  • Run detail: OrchestrationRunDetail shows the live state of a running orchestration
  • Run status badge: RunStatusBadge indicates pending, running, completed, failed, cancelled
  • Run activity feed: RunActivityFeed streams real-time events from the orchestration
  • Run task list: RunTaskList shows tasks with their assigned agents and individual status
  • Run control bar: RunControlBar provides pause, resume, and cancel actions
  • Wave dependency graph: WaveDependencyGraph visualizes the dependency structure between task waves
  • Notifications: useOrchestrationNotifications hook provides real-time alerts for run state changes
  • Agent run card: AgentRunCard shows a single agent’s progress within a run

How to use it

Section titled “How to use it”
  1. Navigate to [Program] > Orchestration (or launch from the Agents page or a Sprint detail).
  2. Click New Run to open the orchestration wizard.
  3. Select the scope — a sprint, workstream, or custom set of tasks.
  4. Choose target repositories for code execution.
  5. Review the AI-generated assignments — adjust agent-to-task mappings if needed.
  6. Preview the plan and click Start to launch the orchestration.
  7. Monitor the run via the activity feed, task list, and dependency graph.
  8. Use the control bar to pause, resume, or cancel the run.

Data model

Section titled “Data model”

This feature uses the following tables:

Section titled “Related”
Agents
Sprints
Tasks