Summary of "Google Antigravity на Русском: Полная Инструкция | Как Создавать мощные ИИ-сервисы и приложения"

Overview

This video is a step-by-step, beginner-friendly guide to Google “Antigravity.” It shows how to install and configure Antigravity, create full applications inside the tool, and link those apps to automation using N8N. The presenter demonstrates building, running, iterating, and deploying real apps from inside Antigravity (Playground → Workspace → repo) and includes live demos: a YouTube video-idea generator and an image-editing web UI connected to N8N.

Installation & Access

Download Antigravity from antigravity.google and install it locally.
Authorization is required. It is recommended to use a Google account set to a US location to avoid regional restrictions and to receive faster updates.
Projects are stored locally in a folder structure similar to a VS Code workspace.

Important Settings & Permissions

Verification policy (MM/agent mode): determine whether the agent should decide when to ask for verification for complex actions or always prompt the user.
Terminal permission: must be granted explicitly if you want Antigravity to run terminal commands or start servers automatically.
Device/security and file/browser permissions: control what Antigravity can access and execute.
Browser integration: Antigravity can open Chrome to preview generated apps.

UI & Work Modes

Workspace vs Playground vs Repositories:
- Playground: sandbox/prototyping environment. Good for experimenting; projects here are not downloadable.
- Workspace: full project environment for production work.
- Repositories: clone from and push to GitHub.
Agent Manager / Agent mode: shows agent activity, allows uploading files/media for context, attaching resources, toggling the terminal, and tracking a checklist of tasks the agent performs.

Model Selection & Run Modes

Recommended models: Google Gemini (Gemini 3 Pro High, Pro Low, Flash). Third‑party models (ChatGPT, Claude) may also appear.
Two run modes:
- Planning mode: builds execution logic, better for complex apps; longer runtime and higher token use.
- Fast mode: quick generation with fewer tokens and less planning.

Typical Development Workflow (Demonstrated)

Create or open a local project folder.
Use Playground to prototype with structured prompts:
- First prompt: high-level plan, architecture, and checklist (data model, UI elements, endpoints).
- Second prompt: detailed implementation requirements, UX guidance, testing and code-quality instructions.
- Optional third prompt: test/fix iteration — ask the agent to run, test, and correct issues.
Agent generates files, creates a server, and exposes a preview link. If the server doesn’t open:
- Ask the agent to start the server, or
- Manually open the Antigravity terminal and run the commands.
Iterate by sending corrective prompts; the agent updates code and restarts as needed.

Integration with N8N (Why & How)

Why use N8N with Antigravity:

Standalone GPT prompts can be inconsistent. Packaging logic into an app and wiring that app to N8N automation produces more predictable, stable outputs.

How it’s set up (overview):

Configure Antigravity MCP (manage MCP reverse): add Antigravity and N8N, then copy relevant config/webhooks from N8N into the Antigravity MCP config.
Import an N8N template (the author provides a template in the pinned comment). Wire webhooks so the Antigravity-generated UI sends images/requests to N8N.
N8N handles the backend flow (for example: send image and prompt to KeyAI → receive generated image → return to UI).

Live demo example:

Upload a source image via the generated UI and add a prompt (e.g., “make the girl hold a tray of oranges”).
The Antigravity UI POSTs to an N8N webhook.
N8N (using KeyAI in the template) returns the generated image and the UI displays it.

Tip: Always include webhook and integration details in your prompt when asking Antigravity to build apps that rely on external automation like N8N.

Deployment & Source Control

Generated projects can be exported to GitHub or deployed to hosting platforms (the demo used Railway after small modifications).
Antigravity supports cloning repositories and pushing generated projects to GitHub.

Practical Tips & Recommendations

Use a US-based Google account/location to avoid region restrictions.
Prefer Google Gemini models for best compatibility with Antigravity.
Use Planning mode for complex, production-grade apps (accept the higher token usage).
Always include webhooks and other integration points in prompts when building apps that rely on external automation.
Keep an N8N template prepared (the author provides one) to ensure automation reliability and consistent results.

Resources & Related Tutorials

N8N template and a separate, detailed video on configuring the template (available via the pinned comment under the video).
A previous video covering MCP / N8N integration details.

Speaker & Tools Shown

Speaker / channel host: Berezuevich
Tools / services demonstrated:
- Google Antigravity (agent + Gemini models)
- N8N
- KeyAI (example for image generation)
- GitHub
- Railway (hosting)
- Local terminal / VS Code–like workspace
- Google Chrome preview