Manager View and Parallel Agents

Manager View is Antigravity's most distinctive feature — and the sharpest departure from traditional IDEs. Where Editor View embeds an agent inside a development surface, Manager View inverts this: the development surfaces are embedded inside the agent.

Google describes it as "a mission control for spawning, orchestrating, and observing multiple agents across multiple workspaces in parallel."

The Paradigm Shift

In Editor View, you are the primary actor and the agent assists. In Manager View, the agent is the primary actor and you supervise.

This matters because the bottleneck in agentic work isn't the model's speed — it's waiting for the human to review each step. Manager View breaks that bottleneck by letting you run multiple agents in parallel while you do other work, and come back to review when they're done.

Spawning an Agent Task

From Manager View, create a new task with a high-level description of what you want done:

Implement the password reset flow: email form, token generation,
and the reset confirmation page. Use the existing auth module
conventions. Write tests.

You don't specify steps. The agent plans the work, breaks it into subtasks, and executes them. As it works, you see a live task list in the Manager with progress indicators.

Running Tasks in Parallel

You can spawn multiple tasks simultaneously — each in its own isolated workspace. For example:

Each workspace is independent. Tasks don't interfere with each other. When a task completes, you review its Artifacts (covered in the next lesson) and decide whether to merge the changes.

Handoffs Between Views

Manager View and Editor View are designed for "instantaneous handoffs." The typical workflow:

1. Manager View — spawn a task, let it run
2. Editor View — agent surfaces a question or needs a decision; you jump in, answer it, hand back
3. Manager View — task resumes, you return to other work

You can also go the other direction: start a task in Editor View, get it to a well-defined state, then hand it off to run autonomously in Manager View while you start something new.

The Async Development Model

Google frames Manager View as forward-looking — a product designed for how development will work as models improve. Today, models can run reliably for stretches of a complex task without intervention. Not yet for days at a time, but the trajectory is clear.

The practical implication now: tasks that would take you 45 minutes of focused coding can run in the background while you're in meetings or working on something else. You return, review the Artifacts, give feedback, and the task continues.

What Manager View Is Not

Manager View is not a way to fire-and-forget work you haven't thought through. Vague tasks produce vague results — the agent still needs a clear objective. The more precisely you describe the goal, the constraints, and the expected output, the better the autonomous run will go.