Playwright MCP (Microsoft)
Official Playwright MCP server: drive a real browser via accessibility snapshots.
Community
What it helps with
Standard tool for 'done means live-verified': click through production after each deploy.
Agent jobs: Venture operator, Any agent
Capability gaps: Deploy & release, Quality verification
Security scan and details
- Scan result
- Vetted pass · 2026-10-17
- Scan notes
- Automated scan: pass (no medium+ findings) at commit b8b4183e099f on 2026-10-17.
- License
- Apache-2.0
- GitHub stars
- 38,280
- Last updated
- 2026-10-16
- Cost
- free
- Source
- github.com/microsoft/playwright-mcp
- Repo
- microsoft/playwright-mcp
Automated static scan of the source at a pinned commit (custom prompt-injection, malware and exfiltration rules plus semgrep, gitleaks and osv-scanner); flagged lines get a manual review. Hosted APIs with no code to scan stay at Needs review. A scan lowers risk; it doesn't prove safety.
Related
Superpowers (obra)
Skills framework/methodology: verification-before-completion, writing-plans, executing-plans, systematic-debugging, TDD, subagent-driven development.
agent-browser (Vercel Labs)
Browser-automation CLI for agents plus its skill (snapshot/click/fill), local or cloud browsers.
Release = verified live (checklist skill)
Don't call a release done until the live URL passes an end-to-end check; keep a rollback ready.
Anthropic Skills (skill-creator, webapp-testing)
Anthropic's public skills repo incl. skill-creator (turn a workflow/correction into a skill) and webapp-testing (Playwright).
Web quality skills (Lighthouse / Core Web Vitals)
Agent skills for auditing and fixing web quality (performance, accessibility, SEO) based on Lighthouse and Core Web Vitals.
Vercel Agent Skills (official)
Official Vercel skills incl. deploy-to-vercel, vercel-cli-with-tokens, React/Next best practices.
Comments
Comments are from visitors, not reviewed by Upskill, and never shown to agents.