Playwright MCP
Microsoft's MCP server that lets LLM agents control a web browser through Playwright.
HN commenters use it to let coding agents like Claude Code drive a browser and take screenshots123, closing the feedback loop so the agent can check its own work.45 The complaints are that it can be slow6 and that it uses more context than alternatives such as Playwriter.7
Why people recommend it
Agent feedback loop · 3
Gives the LLM direct browser feedback so it can validate its work without manual help.
“for web the playwright mcp”
“validate (e.g. using playwright mcp)”
Show 1 more
“now I use playwright mcp for the screenshots and the browser interactions”
Claude Code pairing · 2
Commenters say plain Claude Code with Playwright MCP works well, even beating hyped setups.
“just my vanilla claude code with playwright mcp runs circles around it”
“with playwright-mcp its completely absolved the need for UX designers in my workflow”
Screenshots and interactions · 1
Used for screenshots and browser interactions, replacing an ImageMagick-based comparison step.
“now I use playwright mcp for the screenshots and the browser interactions”
Agent browser access · 1
Recommended as a way to give an agent browser control.
“https://github.com/microsoft/playwright-mcp”
UX reference · 1
One commenter points it at established websites to borrow their UX principles.
“with playwright-mcp its completely absolved the need for UX designers in my workflow”
Caveats and dissent
Speed · 1
One user finds it too slow and says it doesn't understand enough to test fully.
“I use playwright MCP and it's too slow, doesn't understand anything”
Context usage · 1
Playwriter is said to consume less context and be more capable.
“It consumes less context than playwright MCP”
Often compared with
Who recommends it in their own words
Threads where it came up
- Show HN: Make AI coding agents verify its own work 1 mention
- Harness engineering: Leveraging Codex in an agent-first world 1 mention
- Why AI Sucks at Front End 1 mention
- What are the best headless browsers? 1 mention
- Ask HN: Why can't Codex/Claude compile app and test that changes worked? 1 mention
- A macOS app that blurs your screen when you slouch 1 mention
Also written as playwright-mcp. Data: playwright-mcp.json. Quotes are verbatim; the verdict and theme labels are written by a model from these comments, and every clause links to its sources.
All comments 10
“I use playwright MCP and it's too slow, doesn't understand anything”
“validate (e.g. using playwright mcp)”
“now I use playwright mcp for the screenshots and the browser interactions”
“https://github.com/microsoft/playwright-mcp”
“for web the playwright mcp”
“with playwright-mcp its completely absolved the need for UX designers in my workflow”
“Using Playwright MCP a lot”
“just my vanilla claude code with playwright mcp runs circles around it”
“tools like Claude Code and Playwright MCP”
“It consumes less context than playwright MCP”