When Playwright tests fail in CI, you get a trace.zip that's too large for LLM context windows and full of binary data agents can't parse. This server unpacks those archives into 16 focused tools that expose exactly what you need for root cause analysis. You can map locators to source code lines, extract ARIA trees as compressed YAML, filter network logs to only 4xx/5xx responses, correlate DOM mutations with in-flight requests, and pull screenshots from the moment of failure. It handles both local paths and HTTPS URLs, supports pagination to control token costs, and includes a trim operation that strips non-critical screenshots to shrink archives by 70-80%. Reach for this when you're tired of manually copying Playwright logs into ChatGPT and want an agent to diagnose flaky selectors, race conditions, and timing issues automatically.
claude mcp add --transport stdio vola-trebla-playwright-trace-decoder-mcp uvx playwright-trace-decoder-mcp