Documentation

Agent-First Architecture

Antigravity is built around the concept of autonomous agents that can independently handle complex development tasks. Unlike traditional IDEs where AI assists you, Antigravity allows you to delegate entire workflows to agents.

Key Benefits: Parallel task execution, automated testing, continuous learning

Artifacts & Verification

Every agent action generates artifacts - tangible deliverables like task lists, implementation plans, screenshots, and test results. This builds trust and provides transparency into what agents are doing.

Artifact Types: Task lists, screenshots, browser recordings, test reports, code diffs

Multi-Surface Integration

Agents operate across three surfaces: the editor (code), the terminal (commands), and the browser (UI testing). This comprehensive access enables end-to-end feature development.

Capabilities: File editing, command execution, web navigation, DOM inspection

Learning & Knowledge Base

Agents maintain a knowledge base of useful patterns, solutions, and context from past tasks. This allows them to improve over time and apply learned patterns to new situations.

What's Saved: Code snippets, architecture patterns, debugging solutions, API patterns

Feature Development Workflow

1. 1.
   Plan: Describe the feature to an agent. It creates a task list and implementation plan.
2. 2.
   Implement: Agent writes code across multiple files, installs dependencies, and sets up configuration.
3. 3.
   Test: Agent launches the application in the browser and verifies functionality.
4. 4.
   Review: Check artifacts (screenshots, test results) and provide feedback if needed.

Bug Fixing Workflow

1. 1.
   Reproduce: Describe the bug. Agent reproduces it in the browser and captures evidence.
2. 2.
   Investigate: Agent examines related code, logs, and stack traces to identify root cause.
3. 3.
   Fix: Agent implements the fix and creates regression tests.
4. 4.
   Verify: Agent runs tests and confirms the bug is resolved.