ChromeDevTools/chrome-devtools-mcp
Chrome DevTools for coding agents
Read original source
Source record
Recorded metrics
daily_stars · 60 stars today
2026-11-13 14:53:15.571Z
GitHub daily Trending window; attention only
Project analysis
Problem & scenario
The supplied evidence supports Chrome DevTools for agents, with no established connection to WeWorkBuddy. It addresses coding-agent users who need live-browser evidence while debugging. Example scenario, inferred from the documented features, a developer investigates a slow page. Before, the developer manually records a trace in Chrome DevTools and relays findings to the agent. After setup, the agent can open Chrome, record a trace and inspect performance insights directly, reducing manual evidence transfer. Browser content becomes accessible to the MCP client, so the choice of browser session matters when sensitive data is present.
Features & advantages
Chrome DevTools for agents combines browser actions with performance traces, network inspection, screenshots and source-mapped console stacks in an MCP server and CLI. This scope lets a coding agent collect diagnostic evidence alongside browser interactions. Puppeteer supplies automation and automatic waiting for action results, making it a documented dependency. The evidence provides no same-task comparison with a named competing product, so relative reliability or superiority remains unestablished. Official support covers Google Chrome and Chrome for Testing, which constrains browser choice.
Technical approach
The documented performance workflow starts when an MCP client requests a browser-dependent tool. The client launches the server through npx, and the server starts Chrome when the tool needs it. A performance request records a trace and extracts insights for the coding assistant. Requirements include Node.js LTS, npm and supported Chrome. Operationally, connecting the client alone does not start the browser. Performance tools may send trace URLs to the CrUX API, and usage statistics are enabled by default, with separate opt-outs. The supplied README establishes this flow, while commit-pinned implementation details and persistence behavior remain unverified.
Business model & buyers
The documented offering is a package users configure in an MCP client or install globally for CLI access. The supplied material states no price, subscription, payment process or commercial license terms, so it cannot establish a revenue model. Likely evaluators include developers and teams adopting coding agents, with recurring browser-debugging work as an inferred adoption trigger. Agent-tool vendors are another plausible audience because the README explicitly recommends building browser subagents on the project. Any paid support or hosted offering would be hypothetical. Evaluation must account for runtime setup and browser-data access.
Growth & distribution
The clearest documented adoption path runs through existing coding-agent clients, with MCP configuration, CLI installation and plugin routes for Claude Code, Copilot/VS Code and Gemini CLI. First value is concrete, the suggested initial prompt opens Chrome and records a site's performance trace. Included debugging skills could help users extend this into repeated workflows, an adoption hypothesis. Setup requirements and keeping skills aligned with tool versions create friction. S1 records GitHub Trending as a discovery source without a date or rank. Neither that listing nor the README establishes acquisition rates, active use or retention.