Extra
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., "@ExtraOpen Notepad and type a reminder to call mom at 5 PM."
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.
ā” Extra
The Open-Source "Flashless" Computer Use Bridge for Windows 11 & 10
Equip your AI assistants (Antigravity, Claude, AGY, Cursor, ChatGPT) with autonomous, sub-10ms Windows desktop control.
Live Site (extra.yantraos.com) ⢠Architecture ⢠PRD ⢠Roadmap
š What is Extra?
Extra brings the battle-tested computer-use autonomy of yantraOS to Windows 11 and Windows 10 as a lightweight, zero-latency desktop bridge and Model Context Protocol (MCP) server.
While early prototypes like Astra 6 and Anthropic Computer Use rely purely on slow, token-hungry vision loops that misclick on DPI-scaled monitors and get stuck in infinite stalls, Extra introduces a Hybrid Dual-Plane Engine:
ā” Visual Plane: Ultra-fast screen capture (< 3ms) via DirectX Desktop Duplication (DXGI) and pure-ctypes MSS.
šÆ Semantic Plane: Native Windows UI Automation (UIA v3 COM) inspection that queries buttons, menus, and text fields with 100.0% mathematical accuracy.
āØļø Zero-Lag Input: Instant typing via Win32
KEYEVENTF_UNICODE(types 500 characters in 2ms with zero dropped keys and full emoji/multilingual support).š”ļø 2-Strike Stall Breaker: Closed-loop perceptual diffing that detects when an action produces no screen change and prevents infinite click loops.
š 100% Enterprise Clean: Zero obscure binary blobs or solo-dev wheels. Every low-level call relies on Microsoft's built-in Windows system DLLs (
user32.dll,UIAutomationCore.dll,dxgi.dll).
Related MCP server: windows-gui-mcp
ā” 60-Second Quickstart
1. One-Line Install (PowerShell)
Open PowerShell (as regular user) on Windows 10 or 11 and run:
irm https://extra.yantraos.com/install.ps1 | iex2. Connect Your AI Assistant
Extra automatically exposes an MCP Server. Add it to your claude_desktop_config.json or Antigravity MCP settings:
{
"mcpServers": {
"extra": {
"command": "python",
"args": ["-m", "extra.mcp.server"]
}
}
}3. Prompt Your AI
Paste the starter prompt into your AI chat:
"You have access to Extra on my Windows 11 desktop. Open Excel, create a table with the Q3 invoice totals from my Downloads folder, and generate a summary chart."
š Performance Comparison
Metric / Capability | Naive Python ( | Astra 6 / Anthropic Computer Use | Extra "Flashless" Engine |
Screen Grab Speed | 150 ā 300 ms (GDI) | 80 ā 150 ms | 1.8 ā 3.2 ms (DXGI / MSS) |
Typing Speed (100 chars) | 2.5 ā 5.0 seconds | 1.0 ā 2.0 seconds | < 0.005 seconds ( |
Targeting Precision | ~60% (fails on DPI scale) | ~82% (vision heuristic) | 99.4% (Hybrid UIA + DXGI) |
DPI Scaling Support | Broken on 125%/150% | Requires manual client config | Native PerMonitorV2 Auto-Sync |
Token Consumption | High (100% full-screen vision) | High (full-screen vision every step) | 70% lower (UIA tree + ROI crops) |
Infinite Stall Prevention | None (infinite loops) | Basic timeouts | Closed-loop perceptual delta detection |
šļø Repository Structure
extra/
āāā core/ # The Flashless Windows Engine
ā āāā capture.py # Sub-3ms screen capture (DXGI & MSS)
ā āāā geometry.py # PerMonitorV2 DPI scaling & display normalization
ā āāā input_engine.py # Win32 SendInput Unicode & atomic clipboard injection
ā āāā focus.py # AttachThreadInput window focus forcing
ā āāā uia_plane.py # Windows UI Automation v3 COM client
ā āāā stall_breaker.py # Closed-loop perceptual diffing & safety killswitch
āāā fastpath/ # High-speed deterministic execution
ā āāā shell.py # Win32 ShellExecuteEx direct app launcher
ā āāā browser.py # Playwright / Edge CDP DOM bridge
āāā mcp/ # Anthropic Model Context Protocol
ā āāā server.py # Standard JSON-RPC stdio/SSE server
āāā cli.py # CLI runner (extra run, extra doctor, extra test)
āāā install.ps1 # 1-line PowerShell installer
āāā requirements.txt # Enterprise-audited dependency manifest
āāā STARTER_PROMPT.md # Master AI system prompt
āāā PRD.md # Product Requirements Document
āāā ARCHITECTURE.md # Full System Architecture Specification
āāā MEMORY.md # Project Context & Architecture Decision Records (ADRs)š”ļø Security & Enterprise Trust
Extra is engineered from the ground up for strict enterprise compliance:
Zero Solo Binary Blobs: No pre-compiled wheels from unverified repositories that trigger Windows Defender heuristics.
Microsoft & Anthropic Pedigree: Dependencies are strictly limited to official Microsoft (
playwright), Anthropic (mcp), and Python Software Foundation packages.Native System Calls: Low-level OS capabilities utilize Microsoft's pre-installed Windows DLLs directly via Python's standard
ctypes.Zero Telemetry / 100% Sovereign: Extra sends zero telemetry to the cloud. All operations execute strictly between your local AI agent and your local Windows operating system.
š Part of the yantraOS Ecosystem
Extra is the Windows ambassador for yantraOS, the sovereign Arch Linux operating system engineered for autonomous computing.
If you want bare-metal AI autonomy with zero Windows telemetry, sub-microsecond OS kernel scheduling, and native Wayland hardware control, explore yantraOS.
š¤ Contributing & Community
We welcome contributions from kernel hackers, automation researchers, and AI developers! Please read our Contributing Guide and Code of Conduct before submitting pull requests.
For security concerns, please review our Security Policy.
š License
Distributed under the MIT License. See LICENSE for more information.
Copyright (c) 2026 Euryale Ferox Private Limited.
This server cannot be deployed
Maintenance
Related MCP Connectors
Eyes and hands on real Windows PCs ā observe, click, type via Glasswarp API.
- mcp-serverOAuthcom.make
Give your AI agents the tools to build, manage, and run automation workflows.
- openhelmOAuthai.openhelm
Autonomous cloud agent tasks: real browser + your tools, structured evidence-backed results.
Control real Android and iOS devices with LLM agents ā tap, swipe, type, automate flows.
Related MCP Servers
- AlicenseNot gradedqualityBmaintenanceEnables comprehensive Windows desktop automation including screen capture, OCR text extraction, mouse/keyboard control, window management, process control, and clipboard operations through 25+ tools for AI agents.32 PyPI4MIT
- AlicenseNot gradedqualityCmaintenanceEnables AI coding agents to automate Windows desktop applications through semantic UI Automation instead of brittle coordinate clicks, with tools for discovering windows, finding controls by stable identifiers, and verifying actions.1MIT
- AlicenseAqualityAmaintenanceAllows AI clients to see and control Windows 10/11 desktops via MCP, with screenshots, UI Automation, Chrome CDP, keyboard/mouse, and terminal using semantic element targeting.30286 npmMIT
- FlicenseNot gradedqualityCmaintenanceEnables AI agents to control Windows GUI applications like a human using screen capture, OCR, mouse and keyboard input, and window management, with safety levels and memory.-