From qe-framework
Guides setup of Claude-in-Chrome extension for browser automation via MCP. Validates connection status and provides step-by-step configuration instructions.
How this skill is triggered — by the user, by Claude, or both
Slash command
/qe-framework:QchromeThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
This skill is a **setup and usage guide only**. It does NOT execute browser operations.
This skill is a setup and usage guide only. It does NOT execute browser operations.
| Request | Action |
|---|---|
| "Set up Chrome", "Connect browser MCP" | This skill — guide setup |
| "Open a webpage", "Take a screenshot" | NOT this skill — use mcp__claude-in-chrome__* tools directly |
claude mcp list 2>/dev/null | grep -i chrome
Chrome extension that gives Claude Code direct browser control — navigate, click, type, read pages, take screenshots, and record GIFs.
| Feature | Detail |
|---|---|
| Type | Chrome Extension + MCP Server |
| Control | Tab management, navigation, form input, JS execution |
| Output | Page text, screenshots, console logs, network requests |
| Recording | GIF capture of multi-step interactions |
claude mcp add claude-in-chrome -- npx @anthropic/claude-in-chrome-mcp
Or edit ~/.claude.json:
{
"mcpServers": {
"claude-in-chrome": {
"command": "npx",
"args": ["@anthropic/claude-in-chrome-mcp"]
}
}
}
claude mcp list | grep chrome
Restart Claude Code session to activate MCP tools.
In a new Claude Code session, ask Claude to open a tab or read a page to verify the connection works.
| Tool | Description |
|---|---|
tabs_context_mcp | Get info about current browser tabs |
tabs_create_mcp | Create a new tab |
| Tool | Description |
|---|---|
navigate | Navigate to a URL |
computer | Mouse/keyboard control |
find | Find elements on page |
form_input | Fill form fields |
shortcuts_execute | Execute keyboard shortcuts |
shortcuts_list | List available shortcuts |
| Tool | Description |
|---|---|
read_page | Read page content (HTML/accessibility tree) |
get_page_text | Get plain text from page |
read_console_messages | Read browser console output |
read_network_requests | Monitor network activity |
| Tool | Description |
|---|---|
upload_image | Upload an image |
gif_creator | Record browser interactions as GIF |
resize_window | Resize browser window |
| Tool | Description |
|---|---|
javascript_tool | Execute JavaScript in page context |
switch_browser | Switch between browser instances |
update_plan | Update automation plan |
Always call tabs_context_mcp first to get current tab state before creating new tabs.
tabs_context_mcp → navigate (or tabs_create_mcp) → read_page → form_input / computer → verify
gif_creator (start) → capture frames before action → perform action → capture frames after → gif_creator (stop)
Use read_console_messages with pattern parameter for filtered output:
pattern: "[MyApp]" → application-specific logs only
tabs_context_mcp for fresh IDs| Issue | Solution |
|---|---|
| Extension not responding | Check Chrome extension is enabled and active |
| Tab ID invalid | Call tabs_context_mcp to get fresh tab IDs |
| Tools not visible | Restart Claude Code session |
| Dialog blocking | User must manually dismiss alert/confirm dialogs |
| Page not loading | Check URL, network connectivity |
npx claudepluginhub inho-team/qe-framework --plugin qe-frameworkCreates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.