The aiup-playwright skill works without any MCP server — it falls back to your own knowledge and the documentation links inside the skill. The server below lets the agent open the running application in a browser, take an accessibility snapshot, and try a locator before it writes it into a test. It is advisory only; nothing in the skill hard-requires it.

Servers

Server	Type	URL / command	Used by
playwright	stdio	`npx @playwright/mcp@latest`	`playwright-e2e-test` (exploring the UI, debugging tests)

Configure in Claude Code

Add it to your project's .mcp.json (the Tessl tessl mcp start bridge can stay alongside it):

{
  "mcpServers": {
    "playwright": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "@playwright/mcp@latest"
      ]
    }
  }
}

For other agents (Cursor, Gemini, Codex, Copilot), add the same server to that agent's MCP configuration file.

Library documentation

aiup-core's .mcp.json wires up context7 (https://mcp.context7.com/mcp), which resolves current documentation for @playwright/test and any other npm package on demand. With aiup-core installed — a prerequisite for this plugin — no extra documentation server is needed.

Note for Tessl-installed users

Tessl ships it with the plugin: tessl install reads the plugin's bundled .mcp.json and writes it into the MCP configuration of each agent it configures, as the Claude Code marketplace install does. If your agent does not list it after the install — an agent Tessl does not configure, or a manual skill copy — configure the server above manually as shown.