Google Chrome
Install directly from Chrome Web Store for the fastest setup.
onUI gives you two capture modes in one workflow: Annotate mode for element-level fixes and Draw mode for layout or spacing regions. The new compact toolbar design keeps primary actions in a minimal rail, while settings open in a focused pop-out card. With onui-local MCP connected, your agent can read fresh annotations and iterate continuously. No app code changes. No backend.
Install directly from Chrome Web Store for the fastest setup.
Use the Chromium installer path, then load unpacked in Edge.
Use the Firefox installer path, then load temporary add-on.
Watch how annotation-driven AI workflows come together
Element-level and region-level feedback in the redesigned compact toolbar
Run on any supported webpage where content scripts can execute. No app instrumentation required.
Target individual elements, attach intent and severity, and use Shift multi-select when one issue spans multiple components.
Drag rectangle or ellipse regions to call out spacing, alignment, grouping, and layout problems that are bigger than one element.
The redesigned floating rail stays small on-page, while settings open in a readable pop-out card for output level and clear-on-copy controls.
Stable styling that avoids clashes with host app CSS while keeping overlays visually consistent.
Your agent can keep checking annotations via onui_get_report and onui_search_annotations, apply fixes, and repeat in a pair-programming loop.
Your data stays local through native messaging plus a local store. No SaaS dependency required.
You annotate once. Your agent keeps fixing in a loop.
Use Annotate mode for single or multi-element feedback, or Draw mode to mark regions for layout and spacing issues.
Your AI agent reads fresh annotations from local MCP tools like onui_get_report.
Agent applies fixes, you review and add new annotations. Repeat until the UI is production-clean.
Works with Claude Code and Codex as a continuous, annotation-driven UI workflow.
Clear install paths for Chrome, Edge, and Firefox
Chrome installs instantly from Web Store. Edge and Firefox follow the browser-specific setup paths below.
Install directly from Chrome Web Store, then open any page and enable onUI for that tab.
Open Chrome Web StoreRun the installer, then load the unpacked extension from the generated local folder.
macOS/Linux
curl -fsSL https://github.com/onllm-dev/onUI/releases/latest/download/install.sh | bash
Windows PowerShell
irm https://github.com/onllm-dev/onUI/releases/latest/download/install.ps1 | iex
Run the Firefox installer path, then load the temporary add-on using the generated Firefox manifest.
macOS/Linux
curl -fsSL https://github.com/onllm-dev/onUI/releases/latest/download/install.sh | bash -s -- --firefox
Windows PowerShell
iwr https://github.com/onllm-dev/onUI/releases/latest/download/install.ps1 -OutFile install.ps1; .\install.ps1 -Firefox
Every installer path can also set up onui-local. Enter y when prompted if you want the local MCP bridge registered too.
edge://extensions or chrome://extensions~/.onui/extensions/current or %USERPROFILE%\.onui\extensions\currentabout:debugging#/runtime/this-firefox~/.onui/extensions/current/manifest.jsonQuick answers to common questions
onUI now supports Google Chrome, Microsoft Edge, and Mozilla Firefox. Chrome installs from Web Store, while Edge and Firefox are available through unpacked/release-bundle install paths.
onUI is tab-scoped by design. New tabs start with onUI off. This gives you control over which pages have annotation overlays active.
No. The extension, native host, and MCP server all run locally on your machine. Your annotations never leave your device.
Four formats: compact (minimal), standard, detailed, and forensic (maximum context). Choose based on how much information your AI agent needs.