Browser Automation for Browser

AI calls add latency and cost. The natural-language parts are only as reliable as the underlying model. Debugging AI-driven steps is harder than debugging selectors.

It is a control interface, not a complete solution. You still need browser infrastructure (local Chromium or Browserbase). No built-in session management, authentication, or scaling.

chrome-devtools-mcp@1.8.0 engines: "^20.19.0 || ^22.12.0 || >=23". 20.16 and 20.17 fail; 20.19 LTS is in engines. “Node ≥ 22.12” and “Node 20 fails” are both too blunt — README “LTS” is marketing; engines is the catch. A CLI is provided for use without MCP; official CLI docs call it experimental. Complementary to Playwright MCP —DevTools-grade traces, Lighthouse, heap snapshots, and --autoConnect, not a generic scraper or a Playwright replacement. Officially supports Google Chrome and Chrome for Testing only. Exposes browser content to the MCP client—do not point it at sensitive sessions. Usage statistics on by default (opt out with --no-usage-statistics). Needs a current Chrome.

Vs Browserbase (managed fleet), Stagehand (AI Playwright layer), Browser Use (LLM browser framework): Steel is the browser sandbox/API — bring your own agent logic. Self-host means Chrome resource cost; cloud is freemium/paid convenience. Anti-bot arms race still applies.

A live logged-in browser is also a live blast radius: the agent inherits cookies, tabs, permissions, and whatever account is open. Vs Stagehand, Browser Use, and Playwright MCP (separate browser instances or contexts): BrowserSkill trades isolation for continuity. Treat the attached session as the trust boundary, not as a sandbox.

Vs Browser Use: pitch this as the workflow/RPA layer on that lineage, not a second browser agent. AGPL-3.0. README says very early development / not production; optional Browser Use Cloud browsers are convenience, not required for the OSS path.

Vs official Playwright MCP: plain Chromium control — this is stealth Firefox. Vs Browser Use / Steel / Cua: those are agent frameworks or browser APIs; this is the MCP-first stealth layer. Anti-detect purpose → site ToS / abuse risk even though the license is clean.

Anonymous analytics (task metrics and visited domain names) are on by default — turn them off in settings. Chrome and Edge only. It drives your logged-in browser, so a hostile page can try to steer the agent with prompt injection while your sessions are live; use a separate profile for anything sensitive. You pay per token on your own key, and multi-step tasks burn more than you expect. Vs browser-use (Python library you script) and playwright-mcp (browser tools for an agent you already run): Nanobrowser is a point-and-ask extension, not something you build on.

Because it uses your real browser session, an agent can act as you on every site you are logged in to — point it at a separate Chrome profile for anything sensitive. The README still calls it early-stage, and the last commit was January 2026, so momentum has slowed. The extension is installed from GitHub releases, not the Chrome Web Store. Vs chrome-devtools-mcp (official from the Chrome DevTools team): that one gives DevTools-grade traces, Lighthouse and heap snapshots; mcp-chrome is about reusing your logged-in profile for everyday browsing tasks.