Antigravity Dynamic Workflows: How to Use /run-workflow

Last updated: November 14, 2026

Antigravity Dynamic Workflows, added in Google Antigravity App v2.32.0 on November 13, 2026, let you type /run-workflow and a task. The agent writes a short Python script that runs many agents in parallel or in sequence, asks you to approve it and runs it in the background. It needs Google AI Pro or Ultra.

Details below come from Google’s Dynamic Workflows documentation and the Antigravity changelog.

Contents

What changed
How Dynamic Workflows work
How to run a workflow
Script building blocks and options
Permissions and limits
Workflows vs Orchestrator, subagents and /swarm
Frequently asked questions

What changed

The Antigravity App v2.32.0 release notes describe Dynamic Workflows as a way for agents to “orchestrate parallel subagents and multi-step tasks in Python using the Python runtime bundled with the app.” The release also added new themes and prompt navigation shortcuts.

The feature builds on work that first appeared for developers. Antigravity SDK v0.1.22, released October 12, added beta dynamic workflow orchestration with the primitives phase, log, agent, parallel and pipeline. SDK v0.1.23 on October 26 added multi-turn workflow agents. Version 2.32.0 brings the idea to the app as a slash command.

Google’s documentation lists the feature as available on Google AI Pro and Google AI Ultra plans in both the Antigravity App and the Antigravity CLI. New versions roll out gradually over a few days. If you are deciding between plans, see our Google AI Pro vs Ultra comparison.

How Dynamic Workflows work

According to the documentation, each run follows four steps:

The agent reads the built-in run-workflow skill and writes a Python script for your task.
Antigravity shows you the script and asks for approval, unless your permission preset, such as Turbo, already allows it.
The workflow runs in the background, so you can keep working in the same conversation.
When the script finishes, the main agent receives its logs and output and summarizes the results.

Each agent the script starts is a full agent with its own conversation. Agents do not see the script or each other, so the script must give each one everything it needs and ask for a short, well-defined result. Results must match a JSON Schema, which keeps logs and file reads out of your main conversation.

Scripts run in a separate Python process using Python bundled with Antigravity, so you do not need to install it. To use your own interpreter (Python 3.8 or later), set ANTIGRAVITY_PYTHON.

How to run a workflow

Check your plan and version. You need Google AI Pro or Ultra. In the app, Dynamic Workflows arrived in v2.32.0.
Start the command. In a conversation, type /run-work and press Tab to complete /run-workflow, then describe your task. You can also ask in plain words.
Describe items and phases. Google’s example: /run-workflow List the Python files changed since main, then review each file in parallel for unhandled exceptions using the flash model, and log a one-line verdict per file.
Review the script. In the app, choose Yes, allow this time in the “Allow running this workflow script?” prompt. In the CLI, choose Yes, run command. To change the plan, decline and explain.
Follow progress. In the app, expand the workflow step to see phases, logs and agent cards. In the CLI, press Ctrl+O to expand the phase tree, or run /tasks (Ctrl+T) to open Task Details.
Answer agent requests. Workflow agents ask for approval in their own conversations. In the CLI, Ctrl+K approves a pending request.
Read the summary. When the script finishes, the main agent replies with the results. To stop early, stop the parent conversation, which also stops every running workflow agent.

Script building blocks and options

You do not need to write Python, but the building blocks help you review a script. The table summarizes the functions listed in the documentation.

Building block	What it does
`phase(name)`	Starts a named phase; later logs, tool calls and agent cards are grouped under it
`log(*args)`	Adds a line to the current phase and to the result the main agent receives
`agent(prompt, schema, role, type_name, workspace, model)`	Defines an agent; `await agent(...)` runs it once, while a saved agent object can run several turns of one conversation
`parallel(items, fn)`	Runs `fn(item)` for every item at the same time; if one item fails, the others are canceled
`pipeline(items, fn)`	Runs `fn(item)` for one item at a time, in order
`await trajectory(...)`	Returns earlier steps of the current conversation, with optional filters

You can ask for three agent options in your prompt:

Model. Agents inherit your conversation’s model by default. Ask for flash_lite or flash for fast, high-volume checks, or pro for deeper reasoning.
Agent type. self (default), the read-only research agent or a custom subagent.
Workspace. inherit uses your current workspace. branch gives each agent its own Git worktree, which Google suggests when agents edit code in parallel. share gives each agent a new working copy that shares repository storage.

To reuse a script, ask the agent to save it as a .py file, then run /run-workflow Run scripts/review-changes.py.

Permissions and limits

Exact-script approval. You can allow a script once, for the conversation or always. An approval covers that exact script, and a command prefix rule cannot cover a Python script.
Per-agent approvals. Each workflow agent follows your permission preset and waits until you answer its requests.
Fail-fast parallel runs. In parallel(), one failed item cancels the others; use pipeline() for items that share files.
macOS environment. Opened from the Dock, the app does not see shell environment variables such as ANTIGRAVITY_PYTHON.

Workflows vs Orchestrator, subagents and /swarm

Antigravity now has several ways to split work. Google’s comparison describes them as follows.

Mode	Best for
Orchestrator	One conversation where Gemini 4 Carbon or Argon plans and reviews while Gemini 4 Hydrogen executes routine tool calls
Subagents	Delegating one task, such as research or a build
Dynamic Workflows	Scripted control flow: the same check across many files, staged steps, or reviewer-and-fixer loops
`/swarm`	Peer agents coordinating on a shared message board

Orchestrator arrived a day earlier. The Antigravity blog says that on Terminal-Bench 4.1 it reached near-Carbon task success while using 24% less quota on average and finishing 19% faster. Our Gemini 4 Hydrogen guide and Gemini 4 Argon guide cover those models.

If your agents need to talk to each other, Google points you to /swarm, covered in our Antigravity /swarm guide. Terminal users can start with our Antigravity CLI 1.5 guide. A similar idea appears in Anthropic’s tool: our Claude Code Code Mode guide explains how Claude writes JavaScript to batch tool calls, whereas Antigravity’s scripts coordinate whole agents.

Frequently asked questions

What are Dynamic Workflows in Google Antigravity?

Dynamic Workflows are a Google Antigravity feature added in Antigravity App v2.32.0 on November 13, 2026. When you run /run-workflow with a task, the agent writes a short Python script that starts many agents in parallel or in sequence, shows you the script for approval and runs it in the background.

Which plans include Antigravity Dynamic Workflows?

Google’s documentation says Dynamic Workflows are available on Google AI Pro and Google AI Ultra plans, in both the Google Antigravity App and the Antigravity CLI.

Can I choose a cheaper model for workflow agents?

Yes. By default workflow agents inherit your conversation’s model. In your prompt you can ask for a lighter tier such as flash_lite or flash for high-volume checks, or pro for deeper reasoning. An agent fails with the list of valid choices if a tier is not available to you.

How are Dynamic Workflows different from /swarm?

In a workflow, a Python script controls the order of work and agents return structured results without talking to each other. In /swarm, peer agents coordinate on a shared message board.

Can I save a workflow and run it again?

Yes. Ask the agent to save a working script to a .py file, then run it later with /run-workflow Run scripts/review-changes.py. Antigravity asks for approval each time unless you always allow it.

Antigravity Dynamic Workflows: Power-User Guide to /run-workflow, Plans and Scripts