Lens
MCP server for visual feedback, video direction, and QA assertions on web pages, enabling AI agents to read, reply, and resolve annotations in real time.
README
Lens
Point. Describe. Ship.
Chrome extension for visual feedback, video direction, and QA assertions on any web page. Click any element, describe what needs to change, and get structured annotation data that routes directly to your AI coding agent via MCP — or copy it to your clipboard.
Quick Start
npm install
npm run build
Then load dist/ into Chrome:
- Go to
chrome://extensions - Enable Developer mode (top right)
- Click Load unpacked → select the
dist/folder
MCP Setup
Lens ships with an MCP server that lets any AI coding agent in VS Code read your annotations, reply, and resolve them — all through the standard Model Context Protocol.
VS Code (Copilot, Copilot Chat, any MCP-capable extension)
Zero config. Open this project in VS Code. The .vscode/mcp.json file auto-registers the server. Your AI sees the lens_* tools immediately.
Claude Code
Already configured via .claude/settings.json when you open this project. If you want it globally:
{
"mcpServers": {
"lens": {
"command": "npx",
"args": ["tsx", "src/mcp-bridge/server.ts"],
"cwd": "/path/to/lens"
}
}
}
Add to ~/.claude/settings.json.
Cline / Continue / Roo Code
Add to your extension's MCP config (check each extension's docs for the settings file location):
{
"mcpServers": {
"lens": {
"command": "npx",
"args": ["tsx", "src/mcp-bridge/server.ts"],
"cwd": "/path/to/lens"
}
}
}
Cursor
Add to .cursor/mcp.json in your project root:
{
"mcpServers": {
"lens": {
"command": "npx",
"args": ["tsx", "src/mcp-bridge/server.ts"],
"cwd": "/path/to/lens"
}
}
}
Windsurf
Add to ~/.codeium/windsurf/mcp_config.json:
{
"mcpServers": {
"lens": {
"command": "npx",
"args": ["tsx", "src/mcp-bridge/server.ts"],
"cwd": "/path/to/lens"
}
}
}
MCP Tools
Once connected, your AI agent has these tools:
| Tool | Description |
|---|---|
lens_get_pending |
Get all pending annotations (optionally filter by project_slug) |
lens_get_session |
Get a session by ID or URL |
lens_acknowledge |
Mark annotation as seen |
lens_resolve |
Mark annotation as fixed (with optional message) |
lens_dismiss |
Dismiss annotation |
lens_reply |
Reply to an annotation thread |
When the AI resolves or replies, the Chrome extension updates in real time via SSE.
Three Modes
Feedback
Click any element to report visual bugs. Lens captures the DOM selector, bounding box, computed styles, and React/Vue/Angular component tree automatically. You describe what's wrong; Lens packages the context.
Video Director
Build cinematic sequences by selecting elements and assigning camera actions (zoom, pan, highlight, crossfade). Each annotation becomes a sequence step with timing, easing, and optional narration. Export as a Remotion composition manifest.
QA Assertion
Click elements and define what should be true about them (exists, visible, text equals, style matches, layout within bounds). Export as a Playwright test file with real selectors.
Usage
Popup Controls
- Activate Lens — Injects the overlay onto the current page
- Open Panel — Opens the side panel for annotation management and export
- Mode Buttons — Switch between Feedback, Video Director, and QA Assertion
Shortcuts (active when Lens is on)
| Key | Action |
|---|---|
⌘⇧L |
Toggle Lens on/off |
F |
Feedback mode |
V |
Video Director mode |
Q |
QA Assertion mode |
C |
Copy all annotations to clipboard |
H |
Hide/show markers |
Esc |
Deactivate Lens |
1-9 |
Jump to annotation by number |
Side Panel
- View all annotations in the current session
- Set project context (slug, name, repo) for MCP routing
- Export as Markdown, JSON, Playwright tests, or Remotion manifest
- Copy all feedback to clipboard
- Manage annotation status
Clipboard Export
Press C while Lens is active, or use the side panel's Copy button:
# 🔍 Lens Feedback — My App
**Page:** https://example.com/editor
**Date:** Mar 19, 2026
**Items:** 3
## 🔴 Blocking (1)
- **[Fix]** Button overlaps the sidebar on mobile
`<button>` → `div.editor-toolbar > button:nth-child(3)`
Component: EditorToolbar → ActionButton
Architecture
┌─────────────────────────────────────────────────────┐
│ Chrome Extension (Manifest V3) │
│ │
│ ┌──────────────┐ ┌──────────────┐ ┌───────────┐ │
│ │ Content Script│ │ Side Panel │ │ Popup │ │
│ │ • Selection │ │ • Ann. list │ │ • Toggle │ │
│ │ • Highlight │ │ • Export │ │ • Mode │ │
│ │ • Annotation │ │ • Project │ │ • Panel │ │
│ │ popup │ │ context │ │ opener │ │
│ └──────┬───────┘ └──────┬───────┘ └─────┬─────┘ │
│ └─────────┬───────┴────────────────┘ │
│ ┌────────┴────────┐ │
│ │ Service Worker │ │
│ │ • IndexedDB │ │
│ │ • Session mgmt │ │
│ │ • MCP relay │ │
│ └────────┬────────┘ │
└───────────────────┼─────────────────────────────────┘
│
┌────────┴────────┐
│ MCP Server │
│ (stdio + :4848) │
│ • MCP tools │
│ • SSE stream │
└────────┬────────┘
│
┌────────┴────────┐
│ AI Agent │
│ (any MCP client)│
└─────────────────┘
Source Layout
src/
├── background/
│ ├── service-worker.ts # Session persistence, messaging, tab capture
│ └── mcp-server.ts # In-extension MCP tool handler
├── content/
│ ├── content.ts # Main content script entry
│ ├── selector-engine.ts # DOM targeting, CSS/XPath generation
│ ├── annotation-popup.ts # On-page annotation form
│ └── styles.ts # Injected CSS
├── sidepanel/
│ ├── index.html
│ ├── sidepanel.ts # Annotation list, export, project context
│ └── sidepanel.css
├── popup/
│ ├── index.html
│ ├── popup.ts # Mode toggle, side panel opener
│ └── popup.css
├── shared/
│ ├── raf-schema.ts # Annotation format types
│ ├── constants.ts # Design tokens, message types
│ ├── serializers.ts # Markdown, JSON, Remotion, Playwright exporters
│ └── storage.ts # IndexedDB wrapper
├── mcp-bridge/
│ └── server.ts # MCP stdio server + HTTP listener
└── icons/
└── generate-icons.ts # SVG → PNG icon generator
Development
npm run dev # Watch mode
npm run typecheck # Type check
npm run build # Production build
After building, reload the extension in chrome://extensions.
License
MIT
Made with ❤ by RenderDraw
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.