Home / Plugins / browser-harness
browser-harness
browser-use/browser-harnessPlugin
Open-source browser agent driven via CDP — direct browser control, 79K stars, YC W25. browser-harness gives Claude Code direct, low-level control of a real Chrome browser via the Chrome DevTools Protocol. Coordinate-level clicks pass through iframes, shadow DOM, and cross-origin frames at the compositor level — no selector hunting, no Playwright wrappers. Connects to a long-running Browser Use Cloud or local browser session, so logins persist across runs. Highlights: - Screenshot-first workflow: capture → read pixel → click_at_xy → re-capture to verify - Bulk HTTP fetch helper for static pages (no browser overhead) - Raw CDP escape hatch for anything helpers don’t cover - Interaction skills for dialogs, dropdowns, iframes, shadow DOM, file uploads, cross-origin frames - Remote browsers via Browser Use Cloud — share a live URL with a human when login/CAPTCHA blocks the agent
Why this score
Popularity score: 94/100 from public GitHub signals only. It shows how widely used and maintained this is, not how well it works. Not included yet: the fake-star check. The security scan is shown separately and doesn't change the score. How we rank
Security
Scanned Oct 22, 2026 at commit f5eaf904b221, with Cisco's open-source skill-scanner.
What its code does:
- Reads environment variables and sends network requests with them API clients do this to send your key to the service. Check which service it talks to.agent-workspace/domain-skills/browser-use-cloud/cleanup-zombies.py:54 and 2 more
- Runs shell commands Common for dev tools. Read the commands it runs.agent-workspace/domain-skills/browser-use-cloud/cloud.md:1, interaction-skills/connection.md:37, interaction-skills/profile-sync.md:1 and 2 more
- Runs code it builds while running (eval or exec) Harder to read ahead of time, since the code is assembled while it runs.src/browser_harness/run.py:387
- Makes network requestsSKILL.md, agent-workspace/domain-skills/browser-use-cloud/cleanup-zombies.py:48, src/browser_harness/admin.py:525 and 15 more
- Downloads content and runs it Whatever it downloads can change after we scanned it.src/browser_harness/admin.py:363
Web addresses in its code:
- api.browser-use.comagent-workspace/domain-skills/browser-use-cloud/cleanup-zombies.py:33
- browser-use.comagent-workspace/domain-skills/browser-use-cloud/cloud.md:181
- api.manus.imagent-workspace/domain-skills/manus/tasks.md:16
- api.trustpilot.comagent-workspace/domain-skills/trustpilot/scraping.md:254
- booking.comagent-workspace/domain-skills/booking-com/scraping.md:19
- genius.comagent-workspace/domain-skills/genius/scraping.md:109
- indeed.comagent-workspace/domain-skills/job-boards/indeed-glassdoor.md:206
- medium.comagent-workspace/domain-skills/medium/scraping.md:3
- api.github.comagent-workspace/domain-skills/github/scraping.md:152
- claude.aiagent-workspace/domain-skills/claude-ai/extract-share-transcript.py:4
- datadome.coagent-workspace/domain-skills/etsy/scraping.md:18
- developer.etsy.comagent-workspace/domain-skills/etsy/scraping.md:87
- dl.google.comsrc/browser_harness/admin.py:320
- eu.i.posthog.comsrc/browser_harness/telemetry.py:19
- export.arxiv.orgagent-workspace/domain-skills/arxiv-bulk/scraping.md:300
And 15 more.
These are the sites named in the code, so it's where your data could go. It isn't a complete list: code can build addresses while it runs, and a few of these may be links rather than calls. Whatever you send also goes to the AI model you use.
A scan lists patterns in the code. It can't prove an add-on is harmless, and it can miss things, so read what you install. How we scan
Prompts that use Playwright
Alternatives
Other makers' add-ons that do the same job, next to this one.
| Name | Score | Stars | Context | Updated |
|---|---|---|---|---|
| This onebrowser-harness browser-use/browser-harness · Plugin | 94 | ★ 19k | not measured | today |
| Playwright Cloud Browsers Azure/playwright-workspaces · MCP server Also for Playwright | 61 | ★ 15 | not measured | 8d ago |
| browser drobins25/craft · Skill Also for Playwright | 57 | ★ 69 | ≈41 always | 17d ago |
| Web Browser (kogiQA) atagon-GmbH/kogiqa-mcp · MCP server Also for Playwright | 62 | ★ 92 | not measured | 12d ago |
| browser_visible agentscope-ai/skills · Skill Also for Playwright | 55 | ★ 130 | ≈76 always | 2mo ago |
| browser-automation RightNow-AI/openfang · Skill Also for Playwright | 73 | ★ 18k | ≈24 always | 4mo ago |