Meet Antigravity (AGY) — an autonomous, tool-fluent AI coding assistant designed to explore vast codebases, execute terminal workflows safely, dispatch multi-agent swarms, and construct verified living artifacts in real-time.
Antigravity is built on Google DeepMind's cutting-edge Gemini agentic model stack, prioritizing proactive reasoning, rigorous reporting integrity, and precise file surgical patches.
Navigates massive multi-repository workspaces, reads precise line ranges, detects lint issues, and maintains full trajectory memory via structured JSONL transcripts.
Replaces only targeted contiguous blocks with exact pattern verification (`replace_file_content`), eliminating accidental file truncations or phantom refactors.
Separates verified tool results from inferred projections. When blocked or awaiting confirmation, clearly reports exact status without synthetic shortcuts.
Whether you prefer a hyper-responsive terminal interface, an AI-native code editor, a standalone desktop orchestrator, or Python scripts.
agy)Ultra-lightweight, terminal-native agent interface. Fast keyboard shortcuts, interactive slash commands, persistent shell terminals, and reactive non-blocking tasks.
Standalone AI-first development environment featuring interactive inline code lenses, gutter diagnostic diagnostics, real-time split diffs, and conversational sidebars.
Electron-based multi-workspace orchestration center. Complete Auxiliary Pane visualizing subagents, background jobs, live artifacts, and project permission tiers.
Python library for programmatic agent leasing, multi-agent pipeline orchestration, and exposing custom tools via standard Model Context Protocol (MCP).
AGY is equipped with a rich, built-in suite of system inspection, execution, and collaboration tools.
Instantiate specialized child agents (`invoke_subagent`, `define_subagent`) that run asynchronously, report back findings, or collaborate via inter-agent messaging.
Schedule non-blocking recurring cron jobs (`*/5 * * * *`) or delayed one-shot timers with reactive wakeup triggers on specific task completion.
Generate markdown reports with live Mermaid architecture diagrams, tabbed carousels, alert boxes, and interactive approval flows.
Run long builds or persistent daemons with `RunPersistent=true`, stream logs in real-time, send interactive input, and manage tasks asynchronously.
Ultra-fast literal and regex searching across files with glob filters, line-number snippets, and recursive directory mapping.
Read remote web documentation, inspect live browser DOM, and synthesize UI mockups or visual assets using integrated image generation models.
Click any preset or enter custom slash commands to see how Antigravity plans, spawns subagents, calls tools, and writes code.
Supercharge your terminal workflow with specialized autonomous workflows.
| Command | Category | Description & Use Case | Action |
|---|---|---|---|
| /goal <instruction> | Autonomous | Runs long-running or overnight tasks autonomously. The agent performs deep verification and will not stop until the objective is accomplished. | |
| /plan <task> | Planning | Requests a step-by-step technical plan, architectural breakdown, and dependency roadmap before touching any production files. | |
| /schedule <condition/cron> | Background | Sets background timers with condition-aware early wakeups or configures recurring cron schedules for continuous verification. | |
| /grill-me | Alignment | Triggers an interactive technical interview where AGY asks challenging architectural and edge-case questions to refine design decisions. | |
| /teamwork-preview | Swarm | Prepares a multi-agent swarm team layout, distributing specialized research, coding, and testing roles across concurrent subagents. | |
| /learn | Memory | Persists critical user corrections, codebase conventions, and architectural preferences for all future sessions. |
Extend AGY with workspace-specific skills, rule files, sidecar services, and Model Context Protocol servers.
Package reusable scripts, instructions, and templates into modular skills located in `~/.gemini/skills` or `.gemini/skills`.
Enforce coding standards, testing requirements, and security guardrails across every prompt with `.gemini/rules/*.md`.
Connect any external API, database, or tool server natively using the industry standard MCP JSON-RPC protocol.