| name | browser |
| description | Browser automation via the external browsercontainer sidecar (chrome-devtools-mcp). Use for navigating web pages, screenshots, accessibility snapshots, form-filling, JavaScript evaluation, and WebGPU rendering validation. Connects over MCP SSE.
|
| version | 2.0.0 |
| triggers | ["/browser","browse","web automation","scrape","navigate","screenshot"] |
Browser Automation Skill
All browser automation runs on the external browsercontainer sidecar via
chrome-devtools-mcp (40+ tools) over MCP SSE. No local browser is installed
in the agentbox image.
Connection
MCP SSE: http://browsercontainer:8931/sse (chrome-devtools-mcp)
CDP: browsercontainer:9222 (raw Chrome DevTools Protocol)
VNC: localhost:5903 (visual debugging, Display :2)
Auto-registered at boot as browser-gpu in .mcp.json. Manual registration:
claude mcp add browser-gpu --transport sse http://browsercontainer:8931/sse
When To Use
- Page navigation, clicking, typing, form fills
- Screenshots and accessibility snapshots
- JavaScript evaluation in page context
- WebGPU/WebGL rendering validation
- Console log monitoring
- Multi-tab workflows
When Not To Use
- Fetching page content without interaction -- use
web-summary or gemini-url-context
- API testing without a browser -- use
curl or httpx
- Raw CDP protocol scripting -- use the chrome-cdp skill
- QE-grade typed assertions, visual-diff baselines -- use qe-browser
Key Tools
browser_snapshot (preferred for LLM interaction)
Returns an accessibility tree -- structured, deterministic, no vision model needed.
browser_take_screenshot
Capture viewport or full-page screenshot as PNG/JPEG.
browser_navigate
Navigate to a URL.
browser_click / browser_type / browser_fill_form
Interact with page elements.
browser_evaluate
Execute JavaScript in the page context.
browser_console_messages
Read browser console output.
Quick Start
browser_navigate({ url: "https://example.com" })
browser_snapshot()
browser_take_screenshot({ filename: "page.png", fullPage: true })
Sidecar Management
agentbox.sh browsercontainer up
agentbox.sh browsercontainer health
agentbox.sh browsercontainer cdp
agentbox.sh browsercontainer shell
agentbox.sh browsercontainer rebuild
Health Check
curl -s http://browsercontainer:8931/health
curl -s http://browsercontainer:9222/json/list | jq '.[].url'
LaTeX and Diagram Workflows
Mermaid Rendering Workaround
mmdc 11.14.0 in the Nix store has a broken puppeteer dependency. Use the browser sidecar instead to render .mmd files:
mmd_content = pathlib.Path("map.mmd").read_text()
html = f"""<!DOCTYPE html>
<html><head>
<script src="https://cdn.jsdelivr.net/npm/mermaid@11/dist/mermaid.min.js"></script>
<script>mermaid.initialize({{startOnLoad:true,theme:'default'}});</script>
</head><body style="background:white;margin:0;padding:20px">
<div class="mermaid">{mmd_content}</div>
</body></html>"""
html_path = "/tmp/mermaid_render.html"
pathlib.Path(html_path).write_text(html)
Then in the browser sidecar:
browser_navigate({ url: `file://${html_path}` })
browser_evaluate({ expression: "document.querySelector('.mermaid svg') !== null" })
browser_take_screenshot({ filename: "/tmp/wardley_map.png" })
For 2x resolution, set a wider viewport before screenshotting:
browser_evaluate({ expression: "document.body.style.zoom='200%'" })
browser_take_screenshot({ filename: "/tmp/wardley_map_2x.png" })
PDF Preview During LaTeX Builds
Use the sidecar to visually verify LaTeX output during iterative builds:
browser_navigate({ url: "file:///path/to/book/main.pdf" })
browser_take_screenshot({ filename: "/tmp/page_check.png" })
browser_evaluate({ expression: "PDFViewerApplication.page = 42" })
browser_take_screenshot({ filename: "/tmp/page_42.png" })
This is the correct approach for arXiv/book builds — no local PDF viewer required.
