{
  "slug": "control-a-computer",
  "label": "control a desktop computer",
  "blurb": "Agents that see the screen and work the mouse and keyboard on a desktop, the way a person does.",
  "keywords": [
    "computer use",
    "desktop",
    "mouse",
    "screen"
  ],
  "categories": [
    "browser-computer"
  ],
  "access": [
    "api",
    "mcp"
  ],
  "count": 27,
  "official": 10,
  "url": "https://agentswitchboard.dev/tasks/control-a-computer",
  "browseUrl": "https://agentswitchboard.dev/browse?task=control-a-computer",
  "agentsUrl": "https://agentswitchboard.dev/api/agents?task=control-a-computer",
  "agents": [
    {
      "name": "Cua",
      "slug": "cua",
      "url": "https://agentswitchboard.dev/agents/cua",
      "iconUrl": "https://agentswitchboard.dev/icons/cua.png?v=f91a976c",
      "description": "Open-source infrastructure for computer-use agents — virtualized macOS, Linux, and Windows sandboxes, SDKs, and benchmarks for agents that control full desktop environments.",
      "provider": "trycua",
      "categories": [
        "browser-computer"
      ],
      "accessMethods": [
        "api"
      ],
      "verified": true,
      "official": false,
      "featured": false,
      "authType": "apiKey",
      "why": "Desktop Control: Give AI agents full GUI control over desktop apps, browsers, and system interfaces in sandboxes"
    },
    {
      "name": "Screenpipe",
      "slug": "screenpipe",
      "url": "https://agentswitchboard.dev/agents/screenpipe",
      "iconUrl": "https://agentswitchboard.dev/icons/screenpipe.png?v=e29990f2",
      "description": "Records your screen and audio 24/7 locally, then serves that history as searchable context to Claude, Codex and other coding agents so they know what you have seen and done.",
      "provider": "Screenpipe",
      "categories": [
        "memory-state",
        "browser-computer"
      ],
      "accessMethods": [
        "cli",
        "api"
      ],
      "verified": true,
      "official": false,
      "featured": true,
      "authType": "none",
      "why": "Continuous Screen Capture: Records screen and microphone continuously on-device, storing frames and transcripts locally without sending data to a…"
    },
    {
      "name": "UI-TARS Desktop",
      "slug": "ui-tars-desktop",
      "url": "https://agentswitchboard.dev/agents/ui-tars-desktop",
      "iconUrl": "https://agentswitchboard.dev/icons/ui-tars-desktop.png?v=4968d9ba",
      "description": "ByteDance's multimodal GUI agent stack that controls desktop and browser interfaces from natural language using vision-based grounding.",
      "provider": "ByteDance",
      "categories": [
        "browser-computer"
      ],
      "accessMethods": [
        "cli",
        "api"
      ],
      "verified": true,
      "official": true,
      "featured": false,
      "authType": "apiKey",
      "why": null
    },
    {
      "name": "Mac Control MCP",
      "slug": "mac-control-mcp",
      "url": "https://agentswitchboard.dev/agents/mac-control-mcp",
      "iconUrl": "https://agentswitchboard.dev/icons/mac-control-mcp.png?v=91b62e46",
      "description": "Native Swift MCP server with 63 tools for full macOS automation — accessibility tree, Safari/Chrome control, screen capture, OCR, keyboard/mouse, Spotlight, and clipboard.",
      "provider": "AdelElo",
      "categories": [
        "browser-computer"
      ],
      "accessMethods": [
        "mcp"
      ],
      "verified": true,
      "official": false,
      "featured": false,
      "authType": "none",
      "why": "Screen Capture & OCR: Capture screenshots and extract text from any screen region using native macOS OCR engine"
    },
    {
      "name": "Desktop Commander MCP",
      "slug": "desktop-commander-mcp",
      "url": "https://agentswitchboard.dev/agents/desktop-commander-mcp",
      "iconUrl": "https://agentswitchboard.dev/icons/desktop-commander-mcp.png?v=509ce78d",
      "description": "MCP server for full desktop control. Lets AI agents execute shell commands, manage files, edit code with diff patches, and run long processes with live output streaming.",
      "provider": "wonderwhy-er",
      "categories": [
        "browser-computer"
      ],
      "accessMethods": [
        "mcp"
      ],
      "verified": true,
      "official": false,
      "featured": false,
      "authType": "none",
      "why": null
    },
    {
      "name": "Microsoft UFO",
      "slug": "microsoft-ufo",
      "url": "https://agentswitchboard.dev/agents/microsoft-ufo",
      "iconUrl": "https://agentswitchboard.dev/icons/microsoft-ufo.png?v=b809713c",
      "description": "Automates Windows desktop apps through natural-language tasks with the UFO2 Desktop AgentOS, and orchestrates agents across Windows, Linux, and Android devices with UFO3 Galaxy.",
      "provider": "Microsoft",
      "categories": [
        "browser-computer",
        "autonomous-agents"
      ],
      "accessMethods": [
        "cli"
      ],
      "verified": true,
      "official": true,
      "featured": false,
      "authType": "none",
      "why": "Windows Desktop Automation: Completes multi-app Windows tasks from a natural-language request using a HostAgent and AppAgents, launched with python…"
    },
    {
      "name": "Agent S",
      "slug": "agent-s",
      "url": "https://agentswitchboard.dev/agents/agent-s",
      "iconUrl": "https://agentswitchboard.dev/icons/agent-s.png?v=b221f4e5",
      "description": "Operates desktop graphical interfaces like a person, combining hierarchical planning with visual grounding to click through applications and finish multi-step tasks.",
      "provider": "Simular AI",
      "categories": [
        "browser-computer",
        "orchestration"
      ],
      "accessMethods": [
        "api",
        "cli"
      ],
      "verified": true,
      "official": false,
      "featured": false,
      "authType": "apiKey",
      "why": "GUI Grounding: Locates on-screen interface elements from a screenshot and maps them to precise click and type coordinates."
    },
    {
      "name": "Claude Cowork",
      "slug": "claude-cowork",
      "url": "https://agentswitchboard.dev/agents/claude-cowork",
      "iconUrl": "https://agentswitchboard.dev/icons/claude-cowork.png?v=53b7bf2b",
      "description": "Works as an AI coworker in Claude that reads your files, runs scheduled tasks across your apps, and ships finished deliverables back to you. On desktop, web, and mobile. By Anthropic.",
      "provider": "Anthropic",
      "categories": [
        "autonomous-agents",
        "browser-computer"
      ],
      "accessMethods": [
        "api"
      ],
      "verified": true,
      "official": true,
      "featured": false,
      "authType": "oauth2",
      "why": "Computer Use (Dispatch): Operates applications directly via Dispatch computer use to carry out tasks that need real UI interaction."
    },
    {
      "name": "Agent TARS",
      "slug": "agent-tars",
      "url": "https://agentswitchboard.dev/agents/agent-tars",
      "iconUrl": "https://agentswitchboard.dev/icons/agent-tars.png?v=cf57d518",
      "description": "ByteDance's multimodal AI agent for desktop and browser automation — controls GUIs, browsers, and shell via vision and MCP tools. Available as npm CLI and desktop app.",
      "provider": "ByteDance",
      "categories": [
        "browser-computer",
        "infrastructure"
      ],
      "accessMethods": [
        "cli"
      ],
      "verified": true,
      "official": true,
      "featured": false,
      "authType": "none",
      "why": "GUI Control: Control desktop application interfaces using vision-based element detection, clicking, and keyboard input."
    },
    {
      "name": "BitFun",
      "slug": "bitfun",
      "url": "https://agentswitchboard.dev/agents/bitfun",
      "iconUrl": "https://agentswitchboard.dev/icons/bitfun.png?v=d46c9a79",
      "description": "Runs an open-source desktop and terminal coding agent that plans, edits, tests, and reviews changes inside real Git repositories.",
      "provider": "GCWing",
      "categories": [
        "code-devtools",
        "browser-computer"
      ],
      "accessMethods": [
        "cli"
      ],
      "verified": true,
      "official": false,
      "featured": false,
      "authType": "apiKey",
      "why": "Repository Delivery: Plans, edits, tests, and commits changes inside real Git repositories using desktop or terminal interfaces."
    },
    {
      "name": "FIRERPA (lamda)",
      "slug": "firerpa",
      "url": "https://agentswitchboard.dev/agents/firerpa",
      "iconUrl": "https://agentswitchboard.dev/icons/firerpa.png?v=207ce954",
      "description": "Controls Android devices from an on-device server with UI automation, remote desktop, MITM capture, and built-in Frida, exposed to agents through an MCP server and a Python client.",
      "provider": "firerpa",
      "categories": [
        "browser-computer"
      ],
      "accessMethods": [
        "mcp",
        "api"
      ],
      "verified": true,
      "official": false,
      "featured": false,
      "authType": "none",
      "why": "Remote Desktop: Streams and controls the device from a browser with multi-user access."
    },
    {
      "name": "Chrome DevTools MCP",
      "slug": "chrome-devtools-mcp",
      "url": "https://agentswitchboard.dev/agents/chrome-devtools-mcp",
      "iconUrl": "https://agentswitchboard.dev/icons/chrome-devtools-mcp.png?v=28941fdf",
      "description": "Official Chrome DevTools MCP server. Gives AI coding agents browser debugging — inspect DOM, execute JS, capture screenshots, and trace network requests via CDP.",
      "provider": "Chrome DevTools",
      "categories": [
        "browser-computer"
      ],
      "accessMethods": [
        "mcp"
      ],
      "verified": true,
      "official": true,
      "featured": false,
      "authType": "none",
      "why": "Screenshot Capture: Take full-page or viewport screenshots and return them to the agent for visual inspection"
    },
    {
      "name": "Firefox DevTools MCP",
      "slug": "firefox-devtools-mcp",
      "url": "https://agentswitchboard.dev/agents/firefox-devtools-mcp",
      "iconUrl": "https://agentswitchboard.dev/icons/firefox-devtools-mcp.png?v=c0583af3",
      "description": "Lets MCP clients automate and debug a local Firefox via WebDriver BiDi: navigate pages, fill forms, inspect network and console, and take screenshots.",
      "provider": "Mozilla",
      "categories": [
        "browser-computer",
        "code-devtools"
      ],
      "accessMethods": [
        "mcp",
        "cli"
      ],
      "verified": true,
      "official": true,
      "featured": false,
      "authType": "none",
      "why": "Screenshots And Downloads: Captures page screenshots with screenshot_page and tracks or configures file downloads with list_downloads and…"
    },
    {
      "name": "iPhone Use",
      "slug": "iphone-use",
      "url": "https://agentswitchboard.dev/agents/iphone-use",
      "iconUrl": "https://agentswitchboard.dev/icons/iphone-use.png?v=58aa9139",
      "description": "Lets Codex and other coding agents operate a real iPhone over USB through a local MCP server: open apps, read screens, tap, swipe and type. For macOS users with Xcode.",
      "provider": "zhongerxin",
      "categories": [
        "browser-computer"
      ],
      "accessMethods": [
        "mcp"
      ],
      "verified": true,
      "official": false,
      "featured": false,
      "authType": "none",
      "why": "Screen reading: Returns a compact control tree with text and positions, plus real screenshots of the phone."
    },
    {
      "name": "Mobile MCP",
      "slug": "mobile-mcp",
      "url": "https://agentswitchboard.dev/agents/mobile-mcp",
      "iconUrl": "https://agentswitchboard.dev/icons/mobile-mcp.png?v=43c43056",
      "description": "Lets agents automate native iOS and Android apps on simulators, emulators, and real devices through one MCP server, using accessibility trees or screenshots.",
      "provider": "Mobile Next",
      "categories": [
        "browser-computer",
        "code-devtools"
      ],
      "accessMethods": [
        "mcp",
        "cli"
      ],
      "verified": true,
      "official": false,
      "featured": false,
      "authType": "none",
      "why": "Screen Interaction: Lists on-screen UI elements, taps, long-presses, swipes, types text, and presses hardware buttons to drive native app flows."
    },
    {
      "name": "Eigent",
      "slug": "eigent",
      "url": "https://agentswitchboard.dev/agents/eigent",
      "iconUrl": "https://agentswitchboard.dev/icons/eigent.png?v=0f45bba4",
      "description": "Runs a local multi-agent desktop workforce that splits tasks across specialized workers with browser, terminal, and file access on your own machine.",
      "provider": "Eigent AI",
      "categories": [
        "orchestration",
        "browser-computer"
      ],
      "accessMethods": [
        "cli",
        "api"
      ],
      "verified": true,
      "official": false,
      "featured": false,
      "authType": "apiKey",
      "why": null
    },
    {
      "name": "Kampala",
      "slug": "kampala",
      "url": "https://agentswitchboard.dev/agents/kampala",
      "iconUrl": "https://agentswitchboard.dev/icons/kampala.png?v=ad01bb48",
      "description": "YC W26. Watches one human workflow demo and converts any website, mobile, or desktop app into a stable REST API — no browser automation, no selectors, no LLM calls per action.",
      "provider": "Zatanna",
      "categories": [
        "browser-computer"
      ],
      "accessMethods": [
        "api"
      ],
      "verified": true,
      "official": false,
      "featured": false,
      "authType": "apiKey",
      "why": null
    },
    {
      "name": "ARTEMIS",
      "slug": "artemis",
      "url": "https://agentswitchboard.dev/agents/artemis",
      "iconUrl": "https://agentswitchboard.dev/icons/artemis.png?v=d1e5a402",
      "description": "Converts natural-language instructions into reliable Android device automation for coding agents, capturing logs and screenshots from each run.",
      "provider": "Google",
      "categories": [
        "browser-computer",
        "code-devtools",
        "autonomous-agents"
      ],
      "accessMethods": [
        "cli"
      ],
      "verified": true,
      "official": true,
      "featured": false,
      "authType": "none",
      "why": "UI Inspection: Reads the on-screen view hierarchy so an agent can decide what to tap or type next."
    },
    {
      "name": "Ghost OS",
      "slug": "ghost-os",
      "url": "https://agentswitchboard.dev/agents/ghost-os",
      "iconUrl": "https://agentswitchboard.dev/icons/ghost-os.png?v=d027c675",
      "description": "Lets any MCP-capable agent see and operate every macOS app through the accessibility tree, with vision fallback and replayable JSON recipes, for users automating Mac workflows.",
      "provider": "Ghostwright",
      "categories": [
        "browser-computer"
      ],
      "accessMethods": [
        "mcp",
        "cli"
      ],
      "verified": true,
      "official": false,
      "featured": false,
      "authType": "none",
      "why": null
    },
    {
      "name": "Browserless",
      "slug": "browserless",
      "url": "https://agentswitchboard.dev/agents/browserless",
      "iconUrl": "https://agentswitchboard.dev/icons/browserless.png?v=438b2338",
      "description": "Runs hosted or self-hosted headless browsers for agents and scripts, with REST APIs for screenshots, PDFs, and scraping, Puppeteer/Playwright over WebSocket, and an MCP server.",
      "provider": "Browserless",
      "categories": [
        "browser-computer"
      ],
      "accessMethods": [
        "api",
        "mcp",
        "cli"
      ],
      "verified": true,
      "official": false,
      "featured": false,
      "authType": "apiKey",
      "why": "Browser REST APIs: Returns screenshots, PDFs, rendered HTML, or structured scrape results from one-shot HTTP calls to /screenshot, /pdf, /content…"
    },
    {
      "name": "Figma MCP Server",
      "slug": "figma-mcp-server",
      "url": "https://agentswitchboard.dev/agents/figma-mcp-server",
      "iconUrl": "https://agentswitchboard.dev/icons/figma-mcp-server.png?v=65696ce9",
      "description": "Gives coding agents Figma design context, variables, screenshots, and generated code from frames, and lets them write frames and components back to the canvas via remote or desktop MCP.",
      "provider": "Figma",
      "categories": [
        "code-devtools",
        "content-media"
      ],
      "accessMethods": [
        "mcp"
      ],
      "verified": true,
      "official": true,
      "featured": false,
      "authType": "oauth2",
      "why": "Frame Screenshots: Captures a screenshot of the selected node so agents can compare their implementation against the original design."
    },
    {
      "name": "Open Interpreter",
      "slug": "open-interpreter",
      "url": "https://agentswitchboard.dev/agents/open-interpreter",
      "iconUrl": "https://agentswitchboard.dev/icons/open-interpreter.png?v=9ba1b5f8",
      "description": "Lets LLMs run Python, JavaScript, and shell commands locally to summarise PDFs, create charts, control files, and browse the web via a natural language interface.",
      "provider": "Open Interpreter",
      "categories": [
        "browser-computer"
      ],
      "accessMethods": [
        "api",
        "cli"
      ],
      "verified": true,
      "official": false,
      "featured": false,
      "authType": "apiKey",
      "why": "Computer Control: Control the local OS — open apps, navigate UIs, manipulate windows — using multimodal model instructions and keyboard/mouse input."
    },
    {
      "name": "Argent",
      "slug": "argent",
      "url": "https://agentswitchboard.dev/agents/argent",
      "iconUrl": "https://agentswitchboard.dev/icons/argent.png?v=deac3151",
      "description": "Gives AI coding assistants control of iOS simulators, Android emulators, TVs, and Electron apps to tap, debug, profile, and replay flows, for mobile and desktop app developers.",
      "provider": "Software Mansion",
      "categories": [
        "code-devtools",
        "browser-computer"
      ],
      "accessMethods": [
        "mcp",
        "cli"
      ],
      "verified": true,
      "official": false,
      "featured": false,
      "authType": "none",
      "why": null
    },
    {
      "name": "bigarrow (big-arrow-on-the-screen)",
      "slug": "bigarrow",
      "url": "https://agentswitchboard.dev/agents/bigarrow",
      "iconUrl": "https://agentswitchboard.dev/icons/bigarrow.png?v=33a66fc6",
      "description": "Lets coding agents draw arrows, boxes, and labels over any macOS window to show the human what to click, with copy buttons and spoken signs, via a Swift CLI and agent skill.",
      "provider": "Franz Enzenhofer",
      "categories": [
        "productivity",
        "code-devtools"
      ],
      "accessMethods": [
        "cli"
      ],
      "verified": true,
      "official": false,
      "featured": false,
      "authType": "none",
      "why": "Point at UI Elements: Draws an arrow and sign at an element by label, a coordinate, a window, or the mouse, then removes it."
    },
    {
      "name": "Page Agent",
      "slug": "page-agent",
      "url": "https://agentswitchboard.dev/agents/page-agent",
      "iconUrl": "https://agentswitchboard.dev/icons/page-agent.png?v=bf0b467a",
      "description": "Embeds a GUI agent in any web page with one script tag, driving the existing DOM from natural-language instructions without screenshots or a headless browser.",
      "provider": "Alibaba",
      "categories": [
        "browser-computer",
        "code-devtools"
      ],
      "accessMethods": [
        "mcp",
        "browser-extension"
      ],
      "verified": true,
      "official": true,
      "featured": false,
      "authType": "apiKey",
      "why": null
    },
    {
      "name": "Safari MCP Server",
      "slug": "safari-mcp",
      "url": "https://agentswitchboard.dev/agents/safari-mcp",
      "iconUrl": "https://agentswitchboard.dev/icons/safari-mcp.png?v=2d397630",
      "description": "Connects AI agents to a live Safari window for web debugging: DOM inspection, network monitoring, console logs, screenshots, and JS evaluation. Official Apple/WebKit MCP server.",
      "provider": "Apple",
      "categories": [
        "browser-computer",
        "code-devtools"
      ],
      "accessMethods": [
        "mcp"
      ],
      "verified": true,
      "official": true,
      "featured": false,
      "authType": "none",
      "why": null
    },
    {
      "name": "Agent Device",
      "slug": "agent-device",
      "url": "https://agentswitchboard.dev/agents/agent-device",
      "iconUrl": "https://agentswitchboard.dev/icons/agent-device.png?v=4e8b8a78",
      "description": "Automates iOS, Android, HarmonyOS, TV, macOS, and Linux devices from AI coding agents through a CLI, MCP server, and typed Node.js API.",
      "provider": "Callstack",
      "categories": [
        "code-devtools",
        "browser-computer"
      ],
      "accessMethods": [
        "cli",
        "mcp",
        "api"
      ],
      "verified": true,
      "official": false,
      "featured": false,
      "authType": "none",
      "why": "Screen Inspection: Reads the accessibility tree and screenshots so agents can locate elements before acting."
    }
  ],
  "relatedTasks": [
    "automate-a-browser",
    "search-the-web",
    "scrape-websites"
  ],
  "stacks": [
    "iphone-and-android-apps"
  ]
}