Using Han with Google Antigravity

Han works with Google Antigravity through an MCP server bridge that exposes Han's skills, disciplines, and validation hooks as tools the agent can call.

How It Works

Antigravity doesn't have lifecycle hooks like Claude Code or OpenCode. Instead, Han integrates via MCP server, exposing capabilities as tools:

1. MCP tools → Skills, disciplines, validation (agent calls on demand)
2. Native sync → .agent/skills/ and .agent/rules/ (Antigravity discovers natively)
3. Event logging → Browse UI (sessions visible alongside Claude Code sessions)

Agent finishes editing src/app.ts
  -> Agent calls han_validate({ mode: "file", files: ["src/app.ts"] })
  -> Bridge matches PostToolUse hooks (biome, eslint, tsc)
  -> Runs hooks in parallel, returns structured results
  -> Agent fixes validation errors

Setup

1. Install Han and plugins

curl -fsSL https://han.guru/install.sh | bash
han plugin install --auto

2. Add the MCP server to Antigravity

Add to ~/.gemini/antigravity/mcp_config.json:

{
  "mcpServers": {
    "han": {
      "command": "npx",
      "args": ["-y", "antigravity-han-mcp"]
    }
  }
}

3. Optional: Sync skills natively

Ask the agent to run han_sync to copy skills to .agent/skills/ and generate .agent/rules/han-guidelines.md. This lets Antigravity discover Han's skills via its native semantic matching without MCP tool calls.

Available Tools

The bridge exposes five MCP tools:

Tool	Purpose	Usage
han_skills	Browse/load 400+ coding skills	han_skills({ action: "list", filter: "react" })
han_discipline	Activate agent personas (25 disciplines)	han_discipline({ action: "activate", discipline: "frontend" })
han_validate	Run validation hooks on demand	han_validate({ mode: "file", files: ["src/app.ts"] })
han_sync	Sync skills/rules to .agent/ directory	han_sync({})
han_context	Get time, active discipline, stats	han_context({})

Validation

Since Antigravity doesn't have automatic tool lifecycle hooks, validation is on-demand. The agent calls han_validate after edits or before finishing.

Per-File Validation (After Edits)

han_validate({ mode: "file", files: ["src/app.ts"], tool: "Edit" })

Runs matching PostToolUse hooks:

Plugin	What It Does
biome	Lint and format JavaScript/TypeScript
eslint	JavaScript/TypeScript linting
prettier	Code formatting
typescript	Type checking
clippy	Rust linting
pylint	Python linting

Project-Wide Validation (Before Finishing)

han_validate({ mode: "project" })

Runs all Stop hooks for comprehensive project validation:

• Full project linting
• Type checking across the codebase
• Test suite execution

Skills (400+)

The han_skills tool gives the agent access to Han's full skill library without duplicating files:

han_skills({ action: "list", filter: "typescript" })
-> Lists all TypeScript-related skills

han_skills({ action: "load", skill: "typescript-type-system" })
-> Loads full skill content into context

Alternatively, run han_sync to copy skills to .agent/skills/ for native Antigravity discovery via semantic matching.

Disciplines (25 Agent Personas)

Activate specialized agent expertise via han_discipline:

han_discipline({ action: "list" })
-> Available: frontend, backend, sre, security, mobile, database...

han_discipline({ action: "activate", discipline: "frontend" })
-> Agent now operates with frontend specialist context

Available disciplines: frontend, backend, api, architecture, mobile, database, security, infrastructure, sre, performance, accessibility, quality, documentation, project-management, product, data-engineering, machine-learning, and more.

Native Integration via Sync

Running han_sync copies Han's assets to Antigravity's native directories:

• Skills → .agent/skills/han-*/SKILL.md (native semantic discovery)
• Rules → .agent/rules/han-guidelines.md (auto-injected guidelines)

This means Antigravity discovers Han's skills via its built-in semantic matching system, without needing to call MCP tools. The rules file provides core guidelines (professional honesty, no time estimates, skill selection) that are injected into every agent context.

Comparison with Claude Code and OpenCode

Feature	Claude Code	OpenCode	Antigravity
Integration type	Native hooks	JS plugin API	MCP server
Validation trigger	Automatic (PostToolUse)	Automatic (tool.execute.after)	On-demand (han_validate)
Context injection	SessionStart hook	system.transform	.agent/rules/ + han_context
Skill access	han_skills tool	han_skills tool	han_skills + .agent/skills/ sync
Discipline support	do-* plugins	han_discipline tool	han_discipline tool

Remaining Gaps

These are platform limitations in Antigravity (as of Feb 2026):

• Automatic validation: No lifecycle hooks, so validation requires explicit han_validate calls. You can add a rule in .agent/rules/ instructing the agent to validate after edits.
• Permission denial: Cannot block tool execution (no PreToolUse equivalent).
• Subagent hooks: No SubagentStart/SubagentStop equivalent.
• MCP tool events: Cannot intercept MCP tool calls for validation.
• Session checkpoints: Not available.

How Plugins Stay Compatible

Han plugins don't need modification to work with Antigravity. The bridge reads the same han-plugin.yml files that Claude Code and OpenCode use:

# This config works in Claude Code, OpenCode, AND Antigravity
hooks:
  lint-async:
    event: PostToolUse
    command: "npx -y @biomejs/biome check --write ${HAN_FILES}"
    tool_filter: [Edit, Write, NotebookEdit]
    file_filter: ["**/*.{js,jsx,ts,tsx}"]
    dirs_with: ["biome.json"]

Same hook definition. Same validation. Different runtime.

Event Logging

The bridge writes JSONL events to ~/.han/antigravity/projects/. Each event includes provider: "antigravity". The Han coordinator indexes these and serves them through the Browse UI alongside Claude Code and OpenCode sessions.

Next Steps

• Install Han plugins for your project's languages and tools
• Read about hook configuration to understand what hooks do
• Check the bridge source code for implementation details