Lightpanda
Fast headless browser (~107MB binary, Zig). No MCP — direct CLI calls only.
Binary: ~/.local/bin/lightpanda
Core Commands
Fetch a URL (dump HTML)
~/.local/bin/lightpanda fetch --dump https://example.com
Save to file (avoid flooding context)
~/.local/bin/lightpanda fetch --dump https://example.com > /tmp/page.html
With logging
~/.local/bin/lightpanda fetch --dump --log_level info https://example.com
Respect robots.txt
~/.local/bin/lightpanda fetch --dump --obey_robots https://example.com
CDP Server (for Playwright/Puppeteer integration)
~/.local/bin/lightpanda serve --host 127.0.0.1 --port 9222
Then connect Playwright via browserWSEndpoint: 'ws://127.0.0.1:9222'.
Usage Rules
- Always save to /tmp/ when fetching pages — never let raw HTML flood context
- Use
fetch --dumpfor read-only content extraction - Use
serveonly when interactive automation is needed (preferfetchotherwise) - Disable telemetry if needed:
LIGHTPANDA_DISABLE_TELEMETRY=true - Does NOT support screenshots — use Playwright for that
Comparison vs Playwright
| lightpanda | playwright (mcporter) | |
|---|---|---|
| Overhead | None (pure CLI) | MCP + mcporter |
| Use case | Read HTML content | Full browser automation |
| Speed | Instant | Slower startup |
| Screenshots | No | Yes |