接入你的智能体
從發布者安裝。不是 PlugMCP 端點。
遠端 URL 還是本機指令? 兩種安裝方式的差異
Connect Google's cloud AI (Gemini Spark powered by Gemini 3.8 / 3.7) to your local agentic IDE (Antigravity) with 24 powerful MCP tools, hardened by OpenAgentShield Zero-Trust runtime sandboxing and automatic credential redaction.
Quick Start · Architecture · Tools · Connected Apps · Deploy
Gemini Antigravity Bridge is an MCP (Model Context Protocol) server that creates a persistent, bidirectional communication channel between:
Optimized for Gemini 3.8's next-generation agentic reasoning, this enables a fully autonomous loop where Spark can dispatch coding tasks to your local machine, and Antigravity can request cloud intelligence back from Spark — all through a standardized, secure protocol.
┌──────────────────────────────────────────────────────────────────┐
│ GEMINI SPARK (Google Cloud) │
│ 📬 Gmail │ 📁 Drive │ 📅 Calendar │ 🎨 Canva │ 🎥 YouTube │
└─────────────────────────┬────────────────────────────────────────┘
│ MCP over HTTPS
▼
┌──────────────────────────────────────────────────────────────────┐
│ ⚡ GEMINI ANTIGRAVITY BRIDGE (MCP Server) │
│ │
│ 🔧 System Commands 📂 File Operations 🤖 Agent Dispatch │
│ 🧠 Shared Memory 🔄 Project Sync 📊 Status Reports │
│ │
│ Tunnel: ngrok / Cloudflare (permanent HTTPS domain) │
└─────────────────────────┬────────────────────────────────────────┘
│ Local MCP + Message Injection
▼
┌──────────────────────────────────────────────────────────────────┐
│ ANTIGRAVITY IDE (Local Machine) │
│ 🖥️ Terminal │ 📝 Code Editor │ 🌳 Git │ 🧪 Tests │ 🚀 Deploy │
└──────────────────────────────────────────────────────────────────┘
# Clone the repository
git clone https://github.com/nandhakumar-murugan/gemini-antigravity-bridge.git
cd gemini-antigravity-bridge
# Install dependencies or install package directly
pip install -r requirements.txt
# OR install as a CLI tool:
pip install -e .
# Configure environment
cp .env.example .env
# Edit .env with your NGROK_AUTHTOKEN and NGROK_DOMAIN
# Launch the bridge (via script or CLI)
gemini-bridge
# or: python run_with_tunnel.py
https://your-domain.ngrok-free.dev/mcpThrough the bridge, Antigravity can orchestrate Spark's connected Google & third-party apps:
graph TB
subgraph Cloud["☁️ Google Cloud"]
Spark["Gemini Spark<br/>Consumer AI Assistant"]
Gmail["📬 Gmail"]
Drive["📁 Google Drive"]
Calendar["📅 Google Calendar"]
Canva["🎨 Canva"]
end
subgraph Bridge["⚡ Antigravity Bridge"]
MCP["MCP Server<br/>(FastMCP + Starlette)"]
Tunnel["HTTPS Tunnel<br/>(ngrok / Cloudflare)"]
Memory["🧠 Shared Memory<br/>(bridge_history.json)"]
Dashboard["📊 Web Dashboard"]
end
subgraph Local["💻 Local Machine"]
AGY["Antigravity IDE"]
Terminal["🖥️ Terminal"]
Git["🌳 Git Repos"]
Files["📂 File System"]
end
Spark -->|"MCP over HTTPS"| Tunnel
Tunnel --> MCP
MCP -->|"inject_message"| AGY
MCP -->|"run_system_command"| Terminal
MCP -->|"git_quick_status"| Git
MCP -->|"read/write_file"| Files
MCP --> Memory
AGY -->|"save_session_note"| Memory
Memory -->|"get_agent_report"| Spark
Spark --> Gmail
Spark --> Drive
Spark --> Calendar
Spark --> Canva
The bridge includes a silent startup script that launches automatically when you log in:
📁 %APPDATA%\Microsoft\Windows\Start Menu\Programs\Startup\
└── GeminiAntigravityBridge.vbs (silent background launcher)
# Using Docker
docker build -t gemini-antigravity-bridge .
docker run -d -p 8000:8000 --env-file .env gemini-antigravity-bridge
# Using Render.com / Google Cloud Run
# Push to GitHub → Connect repo → Deploy automatically
The bridge supports multi-tier automated monitoring via Gemini Spark schedules:
Contributions are welcome! Please read the Apache 2.0 License before submitting PRs.
git checkout -b feat/amazing-feature)git commit -m 'feat: add amazing feature')git push origin feat/amazing-feature)This project is licensed under the Apache License 2.0 — see the LICENSE file for details.
This framework implements the formal architecture introduced in the published research paper:
AgentShield: A Zero-Trust Runtime Guardrail Architecture for Autonomous Multi-Agent AI Systems with Bidirectional Context Synchronization
Nandhakumar Murugan
Department of Computer Science and Engineering (Cybersecurity), KGiSL Institute of Technology (KiTE), Autonomous
Permanent DOI: 10.5281/zenodo.22259022 | Open Access: Zenodo CERN Repository
If you reference or build upon this architecture in academic or enterprise research, please cite:
@article{murugan2026agentshield,
title={AgentShield: A Zero-Trust Runtime Guardrail Architecture for Autonomous Multi-Agent AI Systems with Bidirectional Context Synchronization},
author={Murugan, Nandhakumar},
journal={Zenodo},
year={2026},
doi={10.5281/zenodo.22259022},
url={https://doi.org/10.5281/zenodo.22259022}
}
Nandhakumar Murugan
B.E. Computer Science & Cyber Security | KGiSL Institute of Technology
Google Student Ambassador (GID: 36) | Open Source Contributor
Built with ❤️ using Google Gemini, DeepMind Antigravity & Model Context Protocol
⚡ Bridge connecting Google Gemini Spark & Cloud AI to your local machine via MCP with autonomous file creation, terminal execution, and subagent orchestration.
工具
尚無工具清單。遠端伺服器會在首次健康檢查後取得工具清單。
常見問題
在本頁複製適用於 Cursor、Claude Desktop、Claude Code 或 VS Code 的 stdio 設定。它使用發布者的安裝方式,而不是 PlugMCP 代理伺服器。
這筆記錄未列出必要的環境變數。
Gemini Antigravity Bridge 透過 stdio 在本機執行。
上次檢查時間:2026-10-28。 ⚡ Bridge connecting Google Gemini Spark & Cloud AI to your local machine via MCP with autonomous file creation, terminal execution, and subagent orchestration.