Universal Local IDE & Workspace MCP Bridge
Provides tools to inspect Git branches, status, and working changes within local projects, and can run git commands via the terminal.
Connects Google Gemini Web (via Connected Apps using StreamableHTTP) to a local IDE or workspace, enabling Gemini to execute tools that read and edit files, run terminal and git commands, and navigate projects.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Universal Local IDE & Workspace MCP BridgeIn my todo-app, add a dark mode toggle and run the tests."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.

Universal Local IDE & Workspace MCP Bridge š
Connect Web AI assistants (ChatGPT, Claude Web) and MCP clients directly and securely to ANY local IDE or developer workspace (VS Code, Cursor, Windsurf, JetBrains / IntelliJ / PyCharm, Antigravity, Zed, Neovim, Terminal) ā with multi-user isolation, zero open ports, and live interactive terminal logs.
š” How It Works
Cloud AI models (ChatGPT, Claude Web) cannot reach into your computer because of firewalls and Wi-Fi NAT.
Universal IDE MCP Bridge creates an on-demand, outbound connection between your local computer and your hosted MCP Gateway:
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā
ā Web AI Assistant ā
ā (ChatGPT / Claude) ā
āāāāāāāāāāāāāāāāāāāā¬āāāāāāāāāāāāāāāāāāāā
ā 1. Tool Call + Your Private Token
ā¼
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā
ā Cloud MCP Gateway (Hosted on Vercel) ā
ā - Authenticates your private token ā
ā - Routes tasks exclusively to your computer ā
ā - Supports Upstash Redis for distributed persistence ā
āāāāāāāāāāāāāāāāāāāā²āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā
ā 2. Outbound polling (Zero inbound ports!)
ā Works behind any Wi-Fi or router
āāāāāāāāāāāāāāāāāāāā“āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā
ā Your Local Computer (On-Demand Bridge Session) ā
ā - Run whenever you want, just like `npm run dev` ā
ā - Live real-time debug logs for every tool & action ā
ā - Reads & edits files, runs terminal/git commands ā
ā - Sandboxed strictly to your chosen projects folder ā
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāRelated MCP server: CodeHands
š 3-Step Setup Guide (ZERO Git Cloning Required)
You do NOT need to clone this GitHub repository to connect your computer. You only need the single standalone bridge.py script.
Step 1: Download bridge.py Globally
Open your terminal in your root projects directory (for example C:\projects on Windows or ~/projects on macOS/Linux), and download the single bridge file:
Windows (PowerShell or Command Prompt):
curl.exe -o bridge.py https://vscode-antigravity-mcp.vercel.app/bridge.pymacOS / Linux:
curl -o bridge.py https://vscode-antigravity-mcp.vercel.app/bridge.py(You can place this single bridge.py file anywhere on your computer ā no additional project files or repositories needed.)
Step 2: Set Your Token & Projects Folder Path
Configure your bridge with your private authentication token and the parent directory containing your code repositories. Using the --save flag saves these settings globally so you only have to do this once:
python bridge.py --save --root "C:\projects" --key "YOUR_SECRET_KEY"--root: The folder containing all your projects (e.g.C:\projectsor/Users/username/projects). Your Web AI will be able to navigate and work across any project inside this folder.--key: Your private auth token (e.g.D4_PROJECT_MCP_SERVERor any custom secret key).--save: Saves this configuration globally to~/.universal_ide_bridge.json.
Verify Your Saved Configuration Anytime:
python bridge.py --show-configOutput:
š Current Saved Global Bridge Configuration:
Config file: C:\Users\Admin\.universal_ide_bridge.json
Server: https://vscode-antigravity-mcp.vercel.app
Root: C:\projects
Token: YOUR_SECRET_KEYStep 3: Run the Bridge (On-Demand Like npm run dev)
Whenever you want to use the AI assistant with your local code, run:
python bridge.py(Or if installed globally, simply run ide-bridge or mcp-bridge)
What Happens When You Run It:
Interactive Console: Launches an active developer session with live status streaming.
Real-Time Debug Logs: Every time ChatGPT or Claude executes a tool, the parameters, execution time (in ms), and output preview are printed in real-time.
Stop Anytime: Press
Ctrl+Cin your terminal to cleanly disconnect and stop the bridge.
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā
ā š Universal IDE Bridge (On-Demand Dev Session) ā
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā
š Projects Root: C:\projects
šÆ Active Workspace: C:\projects\my-app
š Cloud Gateway: https://vscode-antigravity-mcp.vercel.app
š Auth Token: YOUR_SEC..._KEY
š¢ Status: ACTIVE & LISTENING (Press Ctrl+C to stop)
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā
š” Ready! When ChatGPT or Claude calls a tool, logs will stream below:
[21:30:15] š Gateway heartbeat confirmed (latency: 38ms) | Status: Healthy
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā
ā” [21:30:28] INCOMING AI REQUEST: get_active_project
š¦ Parameters: {}
š Target Project: my-app (C:\projects\my-app)
āļø Executing local action...
ā
Local execution completed successfully in 5.8ms
š Result: {"projects_root": "C:\\projects", "active_project": "my-app", "git_branch": "main"}
š Delivered result to AI assistant in 12ms (attempt 1)
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā
š Bridge session stopped by user (Ctrl+C). Clean shutdown complete.
š Goodbye!š Web AI Connection Details (At a Glance)
Connect any Web AI Chat in your browser to your local computer:
Platform | Endpoint URL | Format / Protocol |
ChatGPT Web (Custom GPTs / Actions) |
| OpenAPI Schema Actions (Direct REST) |
ChatGPT Web (MCP Connectors) |
| StreamableHTTP |
Claude Web (Custom Connectors) |
| StreamableHTTP (Official Claude Connector standard) |
Authentication Header |
| Enter your secret key in the connector settings |
URL Query Token Option |
| Append token directly in URL |
š» Browser AI Setup Walkthrough
š¢ 1. ChatGPT Web (chatgpt.com Custom GPT Actions ā Recommended)
In your browser, open chatgpt.com.
Go to Explore GPTs ā + Create (top right) ā Configure tab.
Click Create new action:
Authentication Type:
API KeyāBearerā EnterYOUR_SECRET_KEY.Schema: Click Import from URL ā Enter:
https://vscode-antigravity-mcp.vercel.app/openapi.jsonClick Import. ChatGPT imports all tools immediately!
Click Create / Save ā Select Only me.
Start chatting with your local projects in ChatGPT!
š£ 2. Claude Web (claude.ai Custom Connectors)
Open claude.ai in your browser.
Click your user profile icon (bottom left) ā Settings ā Connectors (or + Add ā Custom connector).
Fill in:
Name:
Universal IDE BridgeURL:
https://vscode-antigravity-mcp.vercel.app/mcp?token=YOUR_SECRET_KEY
Click Continue. Claude connects over official StreamableHTTP and discovers all 17 tools instantly!
š ļø Complete Tools Reference (17 Tools)
š Multi-Project & Workspace Navigation
Tool | Description |
| Discovers and lists all project folders, detected tech stacks, and Git branches across your system. |
| Switches the active project context to a different folder on your machine. |
| Inspects current active project name, workspace path, and Git status. |
| Queries overview of active workspace, Git branch, and working changes. |
š File & Directory Operations
Tool | Description |
| Reads lines or full contents of any file with line numbering. |
| Creates or replaces files in your workspace. |
| Finds and replaces specific code blocks in a file. |
| Creates a new directory (and any necessary parent directories). |
| Deletes a file or directory ( |
| Moves or renames files and directories cleanly. |
| Inspects file size, line count, permissions, extension, and timestamps. |
| Explores directories and inspects the file tree. |
š Code Search
Tool | Description |
| Fast regex/text search for functions, classes, or symbols across your codebase. |
š» Terminals & Version Control
Tool | Description |
| Native PowerShell command execution with full UTF-8 encoding support. |
| Windows Command Prompt ( |
| Direct Git operations ( |
| Default system terminal command execution with configurable timeout. |
āļø CLI Reference (bridge.py)
Flag | Shorthand | Description |
|
| Path to parent directory containing all projects (e.g. |
|
| User Auth Token (default: reads from config or |
|
| Cloud MCP Gateway URL (default: |
| Initial workspace directory (default: current directory) | |
| Saves the provided | |
| Displays current saved global defaults |
āļø For Admins: Deploying Your Own Gateway to Vercel
1. Click Deploy
2. Environment Variables in Vercel Settings
Variable | Description | Example |
| Comma-separated list of |
|
| Secret password for admin API endpoints |
|
| Upstash Redis REST URL for multi-container serverless persistence |
|
| Upstash Redis REST Token |
|
Tip for Vercel Serverless: If you enable free Upstash Redis (
UPSTASH_REDIS_REST_URLandUPSTASH_REDIS_REST_TOKEN), the gateway coordinates tasks across isolated serverless containers with 0% dropped tasks and 0 timeouts.
š Security & Sandboxing
Boundary Protection: The executor enforces strict root directory boundaries. Any attempt by the AI to access files outside your designated root (
../../etc/passwdorC:\Windows) raises an immediatePermissionError.Zero Inbound Ports: Your computer never opens an incoming network port. All communication uses outbound HTTPS polling.
Process Isolation: Terminal commands run inside isolated process groups with Windows
taskkill /F /Ttree-kill timeouts to prevent hanging child processes.
š License
MIT License. Open-source and free for personal and commercial use.
This server cannot be deployed
Maintenance
Related MCP Connectors
The Remote MCP server acts as a standardized bridge between LLM applications (like Claude, ChatGPT, and Cursor) and external services, enabling AI agents to access external tools and resources. Its primary capability is providing a centralized search tool to discover other MCP servers and their respective tools. Unlike local implementations, it runs remotely with OAuth authentication and permission controls for security.
Remote MCP server exposing your Node Command fleet (files, shell, processes, browser) to AI hosts.
The OpenRouter MCP server plugs OpenRouter into the AI tools you already use. Once connected, your assistant can pull live OpenRouter data (models, prices, your credits, rankings, and docs) and send quick test messages, all without leaving your editor.
- QuallaaOAuthcom.quallaa
Talk to your public-facing AI from any MCP client ā Claude, ChatGPT, Cursor, Cline, Windsurf.
Related MCP Servers
- AlicenseNot gradedqualityCmaintenanceLocal MCP server bridging ChatGPT Web to local tools for file, shell, git, test, and process management with secure policy controls.MIT
- FlicenseNot gradedqualityBmaintenanceAn MCP server that gives web AI assistants like ChatGPT and Claude Chat hands-on control over a local development machine, enabling them to read, edit, and execute code through a safe, policy-checked interface.1-
- AlicenseNot gradedqualityBmaintenanceMCP server that lets AI assistants securely operate on local workspaces, including guarded binary and image artifact downloads/uploads on Windows and Linux.3MIT
- AlicenseNot gradedqualityBmaintenanceA self-hosted MCP server that lets ChatGPT securely read, edit, search, and run code in your local projects via an owner-approved connection.547 npmMIT