onUI

onUI

Annotate elements, draw regions, and ship cleaner UI faster.

Category
访问服务器

README

onUI

Annotate Any UI for AI Agents

Lightweight browser extension (Chrome + Edge + Firefox) + local MCP bridge for annotation-first UI pair programming.

Powered by onLLM.dev.

GitHub stars GitHub license Chrome Stable Edge Stable Firefox Stable Buy Me A Coffee

[!NOTE] onUI is now stable and production-ready.

Demo

<a href="https://github.com/onllm-dev/onUI/raw/main/landing-page/assets/onui-demo.mp4"> <img src="https://github.com/onllm-dev/onUI/raw/main/landing-page/assets/onui-demo-preview.gif" alt="onUI demo — click to play full video" width="720"> </a>

<sub>Click the preview above to play the full demo video.</sub>

✨ Why onUI

  • 🧩 No integration into app code
  • 🎛️ Per-tab ON/OFF control (off by default)
  • 🎯 In-page annotation dialog with intent + severity
  • ✍️ Draw mode for region annotations (rectangle + ellipse)
  • ⚙️ Compact toolbar with pop-out settings (output level + clear on copy)
  • 👀 Visual markers and hover targeting
  • 🧾 Export outputs in compact / standard / detailed / forensic formats
  • 🛡️ Shadow DOM isolation for stable styling
  • 🔌 Local MCP server + native bridge (no cloud backend required)

Install (Current)

Option A: Browser Extension Stores (recommended)

Chrome Web Store: https://chromewebstore.google.com/detail/onui/hllgijkdhegkpooopdhbfdjialkhlkan?authuser=0&hl=en-GB

Microsoft Edge Add-ons: https://microsoftedge.microsoft.com/addons/detail/onui/fkcmlckehjhcicihbnmhkadfhjhfnond

Option B: One-command installer from latest GitHub release

Use this if you want the latest version (store updates may lag) or for Firefox:

macOS/Linux:

curl -fsSL https://github.com/onllm-dev/onUI/releases/latest/download/install.sh | bash

macOS/Linux (Firefox artifact):

curl -fsSL https://github.com/onllm-dev/onUI/releases/latest/download/install.sh | bash -s -- --firefox

Windows (PowerShell):

irm https://github.com/onllm-dev/onUI/releases/latest/download/install.ps1 | iex

Windows (PowerShell, Firefox artifact):

iwr https://github.com/onllm-dev/onUI/releases/latest/download/install.ps1 -OutFile install.ps1; .\install.ps1 -Firefox

The installer handles extension install and can set up MCP in the same run. When prompted with Set up local MCP bridge now? [y/N], enter y to enable MCP.

Then load it in Chrome or Edge:

  1. Open chrome://extensions or edge://extensions
  2. Enable Developer mode
  3. Click Load unpacked
  4. Select ~/.onui/extensions/current (or %USERPROFILE%\.onui\extensions\current on Windows)

Chromium browsers require this final manual step for unpacked extensions.

Firefox (manual from release artifact):

  1. Download onui-firefox-add-ons-vX.Y.Z.zip from the GitHub release page.
  2. Extract it to a local folder.
  3. Open about:debugging#/runtime/this-firefox.
  4. Click Load Temporary Add-on...
  5. Select the extracted manifest.json.

🧠 Usage

  1. Open any supported website tab.
  2. Click the onUI extension icon.
  3. Enable This Tab.
  4. Use the on-page launcher to open the compact toolbar.
  5. Toggle Annotate mode for element targeting or Draw mode for region targeting.
  6. Hold Shift and click multiple elements to batch-select targets.
  7. Release Shift to open a shared annotation dialog for selected targets.
  8. Save once to create one annotation per selected element (or one region annotation in draw flow).
  9. Open toolbar Settings to choose output level and configure Clear on copy.
  10. Copy exported output from the toolbar.

🔌 Local MCP Setup

onllm-dev/onUI MCP server

Recommended path: use the same installer command above and answer y when prompted.

If you want to force MCP setup in non-interactive mode:

macOS/Linux (--mcp):

curl -fsSL https://github.com/onllm-dev/onUI/releases/latest/download/install.sh | bash -s -- --mcp

Windows (PowerShell, set env var before running installer):

$env:ONUI_INSTALL_MCP=1; irm https://github.com/onllm-dev/onUI/releases/latest/download/install.ps1 | iex

MCP setup now uses a prebuilt release bundle (no local build required), but still needs Node 20+.

Manual MCP setup from source is still supported:

pnpm build:mcp
pnpm setup:mcp
pnpm doctor:mcp

Manual JSON config for custom MCP routers/clients

If your MCP router uses an object-style mcpServers map, use this canonical entry:

{
  "mcpServers": {
    "onui-local": {
      "command": "node",
      "args": [
        "/ABSOLUTE/PATH/TO/onUI/packages/mcp-server/dist/bin/onui-cli.js",
        "mcp"
      ]
    }
  }
}

Use an absolute path for onui-cli.js (relative paths are often rejected or resolved incorrectly by routers).

If your router uses a list/array schema instead of an object map, adapt the same command/args shape like this:

{
  "servers": [
    {
      "name": "onui-local",
      "command": "node",
      "args": [
        "/ABSOLUTE/PATH/TO/onUI/packages/mcp-server/dist/bin/onui-cli.js",
        "mcp"
      ]
    }
  ]
}

The list example above is a schema adaptation pattern, not a claim about any specific router's exact key names.

Setup/verification notes:

  • Run pnpm build:mcp first so packages/mcp-server/dist/bin/onui-cli.js exists.

  • Keep the server entry name as onui-local.

  • Run pnpm doctor:mcp after wiring config to confirm local setup health.

  • Auto-registers onui-local for Claude Code and Codex when those CLIs are installed.

  • Browser support in this release: Chrome stable + Edge stable + Firefox stable (unpacked).

  • @onui/mcp-server is workspace-local (private: true), so run setup/doctor from this repo.

See:

  • docs/mcp-setup.md
  • docs/doctor.md
  • docs/release.md

Maintainer Build + Release

app.sh is the local release entrypoint (no CI/CD dependency).

Local validation + artifact packaging

./app.sh --build

This runs:

  1. Prereq checks (Node 20+, pnpm, git, zip)
  2. Build order: @onui/core -> @onui/extension -> @onui/mcp-server
  3. MCP tests
  4. MCP doctor smoke check (warnings allowed, errors fail)
  5. Artifact packaging into artifacts/vX.Y.Z/

Artifacts:

  1. onui-extension-unpacked-vX.Y.Z.zip
  2. onui-chrome-web-store-vX.Y.Z.zip (manifest key stripped for CWS)
  3. onui-edge-add-ons-vX.Y.Z.zip (manifest key stripped for Edge Add-ons)
  4. onui-firefox-add-ons-vX.Y.Z.zip
  5. onui-mcp-bundle-vX.Y.Z.zip
  6. install.sh
  7. install.ps1
  8. checksums.txt

Local release + GitHub publish

./app.sh --release

Release gates:

  1. Clean git tree
  2. Current branch is main
  3. gh auth status succeeds

Release actions:

  1. Auto patch bump from root package.json
  2. Sync version across extension + MCP runtime strings
  3. Run full --build
  4. Commit + tag vX.Y.Z
  5. Push commit/tag
  6. Create GitHub release with packaged assets

🛠️ Development

pnpm install
pnpm check
pnpm test:coverage

🗂️ Repository Structure

packages/
  core/        Shared annotation/report types + formatters
  extension/   Browser extension runtime (background/content/popup)
  mcp-server/  Local MCP server + native bridge setup/doctor tooling

⭐ Support

If onUI is useful to you, please star the repo: https://github.com/onllm-dev/onUI

It helps other users discover the product.

<a href="https://buymeacoffee.com/tushar_s" target="_blank"><img src="https://cdn.buymeacoffee.com/buttons/v2/default-yellow.png" alt="Buy Me A Coffee" height="48"></a>

Star History

Star History Chart

📄 License

GPL-3.0

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选