\r\n\r\nUindow drives a **real, signed Chromium/Electron browser** with genuine OS-level\r\ninput - actual cursor movement, real keystrokes, and native file dialogs rather\r\nthan synthetic page events. It runs entirely on your own machine and your own\r\nnetwork, and every line of code it executes sits in plain sight in this repository.\r\n\r\nAutomate it three ways:\r\n\r\n- **No code** - build automations in the integrated editor and record complex workflows without writing code.\r\n- **From any AI assistants** - Uindow ships a local\r\n [MCP](https://modelcontextprotocol.io) server, so Claude, Cursor, VS Code, or any\r\n MCP-compatible client can list, create, and run automation modules directly.\r\n See [Control Uindow from AI assistants](#control-uindow-from-ai-assistents-mcp).\r\n- **Write automations in pure JavaScript** - Use the integrated development environment to write, test, and\r\n debug automations with ease. Auto-completion, code healing, JavaScript parsing, and linting are all built in.\r\n\r\n## Running Uindow\r\n\r\nIn order to use the app, create an account for free at [uindow.com](https://uindow.com/?ref=github)\r\nand follow the on-screen instructions.\r\n\r\n### Option 1 - Prebuilt binaries (recommended)\r\n\r\nWe build signed binaries for **macOS, Windows, and Linux** directly from the `dist`\r\nsource, and host them on the [GitHub Releases](https://github.com/uindow/uindow/releases)\r\npage (current and older versions) and on the [Uindow install page](https://uindow.com/install/).\r\n\r\n### Option 2 - `npx`\r\n\r\nOne command to fetch the CLI and launch the app:\r\n\r\n```bash\r\nnpx -y @uindow/cli@latest app:start\r\n```\r\n\r\nOther lifecycle commands:\r\n\r\n```bash\r\nnpx @uindow/cli app:status # check whether the app is running\r\nnpx @uindow/cli app:stop # stop the app\r\n```\r\n\r\n### Option 3 - Run from source\r\n\r\nClone the repository and launch the app directly from source.\r\n\r\n```bash\r\ngit clone https://github.com/uindow/uindow.git uindow\r\ncd ./uindow/\r\nnpm install\r\nnpm start\r\n```\r\n\r\n## Control Uindow from AI assistents (MCP)\r\n\r\nUindow exposes a local [Model Context Protocol](https://modelcontextprotocol.io) server,\r\nso any MCP-compatible client can drive web-automation agents directly. \r\nThe server communicates over **stdio** securely and is launched on demand by the AI assistant.\r\n\r\n**Requirements:** Node.js 18+ with `npx` on your `PATH`.\r\n\r\n```bash\r\nnpx -y @uindow/cli@latest mcp\r\n```\r\n\r\n### Quick reference\r\n\r\nUindow provides a 1-click connector for the most popular AI assistants.\r\n\r\n> Go to **Uindow** > **AI assistants** > **Connect**\r\n\r\n| AI assistant | Instant connect | Root key | Config location |\r\n| --- | --- | --- | --- |\r\n| Claude Code (CLI + Desktop **Code**) | ✅ yes | `mcpServers` | `~/.claude.json` |\r\n| Claude Desktop (**Chat** / **Cowork**) | ✅ yes | `mcpServers` | `claude_desktop_config.json` |\r\n| Cursor | ✅ yes | `mcpServers` | `~/.cursor/mcp.json` |\r\n| VS Code (Copilot) | ✅ yes | `servers` | `mcp.json` (**MCP: Open User Configuration**) |\r\n| Windsurf | ✅ yes | `mcpServers` | `~/.codeium/windsurf/mcp_config.json` |\r\n| Zed | ✅ yes | `context_servers` | `settings.json` |\r\n| Codex | ✅ yes | **TOML** `[mcp_servers.*]` | `~/.codex/config.toml` |\r\n| Gemini CLI | ✅ yes | `mcpServers` | `~/.gemini/settings.json` |\r\n| Cline | ✅ yes | `mcpServers` | `cline_mcp_settings.json` |\r\n| Goose | ✅ yes | **YAML** `extensions` | `~/.config/goose/config.yaml` |\r\n\r\n---\r\n\r\n### JetBrains AI Assistant\r\n\r\nGo to **Settings > Tools > AI Assistant > Model Context Protocol (MCP)** and click\r\n**Add**, then paste the JSON. In the same dialog, set the scope to **Global** rather than\r\nproject-scoped so the server is available in every project you open, then click **Apply**\r\nto start it.\r\n\r\n```json\r\n{\r\n \"mcpServers\": {\r\n \"uindow\": {\r\n \"command\": \"npx\",\r\n \"args\": [\"-y\", \"@uindow/cli@latest\", \"mcp\"]\r\n }\r\n }\r\n}\r\n```\r\n\r\n### Any other AI Assistant\r\n\r\nMost remaining AI assistants accept the same object under `mcpServers`. Look for the config file\r\nin your home directory (`~/./...`) rather than the one in your project root - the\r\nhome-directory copy is the global one:\r\n\r\n```json\r\n{\r\n \"mcpServers\": {\r\n \"uindow\": {\r\n \"command\": \"npx\",\r\n \"args\": [\"-y\", \"@uindow/cli@latest\", \"mcp\"]\r\n }\r\n }\r\n}\r\n```\r\n\r\nPoint the AI assistant at the command `npx` with arguments `-y @uindow/cli@latest mcp`.\r\n\r\nIf your AI assistant requires an explicit transport field, use `\"type\": \"stdio\"`.\r\n\r\n### Verifying the connection\r\n\r\nRun the launch command by hand first - it's the fastest way to separate \"Uindow is broken\"\r\nfrom \"the AI assistant can't start it\":\r\n\r\n```bash\r\nnpx -y @uindow/cli@latest mcp\r\n```\r\n\r\nA stdio server prints nothing and blocks. A silent, hung terminal means it started\r\ncorrectly; press Ctrl-C and go back to your AI assistant. Any stack trace you see here is the\r\nreal error your AI assistant was swallowing.\r\n\r\nThen confirm the scope took: open the AI assistant from a **different** directory than the one\r\nyou configured it in, and check that Uindow's tools are still listed. If they vanish, the\r\nentry landed in a project-local config.\r\n\r\n#### Troubleshooting\r\n\r\n* **Server works in one project but not another.**\r\n * Classic scope problem - the entry is\r\n project-local. In Claude Code, `claude mcp list` from the other directory will come up\r\n empty; re-add with `--scope user`. In Gemini CLI, re-add with `-s user`. In Cursor and VS\r\n Code, move the entry from `.cursor/mcp.json` or `.vscode/mcp.json` to `~/.cursor/mcp.json`\r\n or the user-profile `mcp.json`.\r\n\r\n* **`spawn npx ENOENT` / server never starts in a GUI app.** \r\n * Desktop apps don't inherit your\r\n shell's `PATH`, which bites anyone using nvm, asdf, or Volta. Run `which npx` (`where npx`\r\n on Windows) and put the absolute path in `command`.\r\n\r\n* **Config saved, nothing happened.** \r\n * Most desktop AI assistants only reload MCP config on a full\r\n restart - quit the app entirely (macOS: Cmd+Q; Windows: quit from the tray icon) rather\r\n than closing the window.\r\n\r\n* **Tools missing after adding the server.** \r\n * Check the root key against the table above:\r\n `servers` for VS Code, `context_servers` for Zed, `mcp_servers` in TOML for Codex,\r\n `mcpServers` everywhere else. A wrong key is ignored silently in most AI assistants.\r\n\r\n* **Duplicate or shadowed entries.** \r\n * Several AI assistants resolve project config ahead of global\r\n config, so an old project-local `uindow` entry will silently win over the new global one.\r\n Delete the stale entry rather than editing both.\r\n\r\n* **Server connects but tools aren't used.** \r\n * In Claude Code, MCP tools are deferred behind\r\n tool search by default and loaded on demand, so they may not appear in an upfront tool\r\n list. Ask for a Uindow tool by name, or set `\"alwaysLoad\": true` on the server entry to\r\n load its tools at session start.\r\n\r\n* **Where to look next.** \r\n * `claude mcp list` and `/mcp` (Claude Code), Output panel > MCP\r\n Logs (Cursor), Output > MCP (VS Code), `~/Library/Logs/Claude/mcp*.log` or\r\n `%APPDATA%\\Claude\\logs\\mcp*.log` (Claude Desktop).\r\n\r\n\r\n## Command-line interface\r\n\r\nYou can run Uindow from any CI/CD pipeline or command-line interface.\r\n\r\n```bash\r\nnpx -y @uindow/cli@latest --help\r\n```\r\n\r\nAlternatively, you can use `node dist/bin.js --help` instead of `npx @uindow/cli --help`\r\nfor a faster response.\r\n\r\n```\r\n USAGE\r\n $ npx @uindow/cli [options]\r\n\r\n AVAILABLE COMMANDS\r\n $ npx @uindow/cli mcp Uindow MCP server (stdio) \r\n $ npx @uindow/cli app_docs Query SDK documentation \r\n $ npx @uindow/cli app_status Get application status \r\n $ npx @uindow/cli app_start Start application \r\n $ npx @uindow/cli app_stop Stop application \r\n $ npx @uindow/cli agent_list List agents \r\n $ npx @uindow/cli agent_create Create agent \r\n $ npx @uindow/cli agent_update Update agent \r\n $ npx @uindow/cli agent_delete Delete agent \r\n $ npx @uindow/cli agent_status Get agent status \r\n $ npx @uindow/cli agent_start Start agent \r\n $ npx @uindow/cli agent_stop Stop agent \r\n $ npx @uindow/cli agent_screenshot Grab agent screenshot \r\n $ npx @uindow/cli agent_viewport Scan agent viewport \r\n $ npx @uindow/cli agent_execute Execute code in an agent \r\n $ npx @uindow/cli agent_logs Get agent logs \r\n $ npx @uindow/cli agent_settings_get Get settings \r\n $ npx @uindow/cli agent_settings_set Update settings \r\n $ npx @uindow/cli agent_results_get Get latest results \r\n $ npx @uindow/cli src_sample Module: get sample (yaml) \r\n $ npx @uindow/cli src_export Module: get source (yaml) \r\n $ npx @uindow/cli src_code_get Module: get state or function code \r\n $ npx @uindow/cli src_code_set Module: set state or function code \r\n $ npx @uindow/cli src_keys_list Module: list item keys \r\n $ npx @uindow/cli src_keys_reorder Module: reorder item keys \r\n $ npx @uindow/cli src_keys_update Module: update item key \r\n $ npx @uindow/cli src_keys_delete Module: delete item \r\n $ npx @uindow/cli src_input_get Module: get input \r\n $ npx @uindow/cli src_input_set_bool Module: set boolean input \r\n $ npx @uindow/cli src_input_set_int Module: set integer input \r\n $ npx @uindow/cli src_input_set_string Module: set string input \r\n $ npx @uindow/cli src_input_set_files Module: set files input \r\n $ npx @uindow/cli src_input_set_table Module: set table input \r\n $ npx @uindow/cli src_output_get Module: get output \r\n $ npx @uindow/cli src_output_set_bool Module: set boolean output \r\n $ npx @uindow/cli src_output_set_int Module: set integer output \r\n $ npx @uindow/cli src_output_set_string Module: set string output \r\n $ npx @uindow/cli src_output_set_files Module: set files output \r\n $ npx @uindow/cli src_output_set_table Module: set table output \r\n\r\n OPTIONS\r\n --help Help menu for a specific command \r\n --version Package version \r\n```\r\n\r\n## Creating modules\r\n\r\nThere are three ways to build a module - reach for them in this order:\r\n\r\n1. **Record it - zero learning curve.** Open the integrated recorder and use the\r\n browser exactly as you normally would: point, click, scroll, upload and download\r\n files. The recorder turns your actions into JavaScript for you - deterministically,\r\n without any AI, and instantly. What you see is what you get.\r\n\r\n2. **Let an AI assistant write it via MCP.** Want something more involved? Hand control to\r\n your local AI assistant over MCP and have it author the module on your behalf. Describe\r\n the outcome and let it produce the code for you. You're always in control of your\r\n automations, and you can use the included IDE to debug your code.\r\n\r\n3. **Write it yourself in plain JavaScript.** If the recorder and the AI-driven approach \r\n both come up short, go straight to the source:\r\n 1. Visit the [Uindow SDK Reference](https://uindow.com/docs/)\r\n 2. Download the sample module and import it into Uindow\r\n 3. Experiment with the dollar-sign methods - the integrated editor has auto-complete,\r\n code hints, formatting and linting\r\n\r\nFor most people the learning curve is zero - the recorder is all you'll ever touch.\r\nAnd if you decide to go pro, it stays shallow: the SDK is there when you want it, not\r\nbefore.\r\n",
"bytes": 11857,
"sha": "f3cb6fca9e75b8d0da3b6bbdd0529929a2d5c7c4bb86b9cad205e33b936f33c8",
"repo_slug": "uindow/uindow",
"fonte": "repo",
"truncated": false,
"api": "https://agentalog.com/api/listings/mcp_io_github_uindow_uindow_bd71b177/readme"
}
![](\"https://uindow.github.io/img/github-banner.png\"/)
\r\n \r\n