playwright-min-network-mcp

Intercepts HTTP/HTTPS network requests made during Playwright browser automation by hooking into the browser's network event stream, capturing request metadata (URL, method, headers, body) and response data (status, headers, body) without modifying page behavior. Uses Playwright's built-in request/response event listeners to create a minimal logging pipeline that streams network activity to the MCP client for real-time inspection.

Captures and stores the full response body content (HTML, JSON, binary data) for each network request, using Playwright's response.body() or response.text() methods to extract payloads after the response is received. Implements optional filtering to exclude large binary responses (images, videos) and provides structured access to response content for assertion and analysis.

Filters network events based on configurable criteria (URL patterns, HTTP methods, content-type headers, domain whitelist/blacklist) to reduce noise and focus monitoring on relevant traffic. Implements pattern matching using regex or glob syntax to route different request types to different handlers or storage backends, enabling selective logging without capturing all network activity.

Extracts timing metrics from network requests including request duration, time-to-first-byte (TTFB), DNS lookup time, and connection establishment time using Playwright's request/response timing data and HAR-compatible timing objects. Aggregates metrics across requests to compute summary statistics (average, p95, p99 latency) for performance analysis and bottleneck identification.

Exposes network monitoring capabilities as MCP tools and resources, allowing LLM clients to subscribe to real-time network events, query historical network logs, and trigger network monitoring on-demand. Implements MCP resource endpoints for accessing captured network data and tool endpoints for controlling monitoring behavior (start, stop, filter, export), using stdio transport for communication with LLM agents.

Detects and categorizes network failures including failed requests (4xx, 5xx status codes), connection errors, timeouts, and protocol violations by analyzing response status codes and error events. Provides structured error metadata (error type, status code, error message) and enables filtering to focus on failure scenarios for debugging and test assertions.