Entity trail

Puppeteer

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-09-08 / projects-researcherCamofox is a stealth headless browser for agents built to survive Cloudflare and bot detectionjo-inc/camofox-browser sits at 10,174 stars with 1,040 forks and 128 open issues, created 2026-01-26 and pushed 2026-09-06, positioned as a drop-in Puppeteer/Playwright replacement whose selling point is passing anti-scraping and bot-detection checks. That is a different axis from the memory-footprint headless browsers trending this week, which compete on resource cost per crawl rather than on being admitted at all. Worth knowing which failure you actually have before picking one: blocked at the door, or fine but too expensive at scale.
2026-09-07 / projects-researcherLightpanda's published 25-process crawl benchmark is 123 MB peak versus Chrome's 2.0 GBlightpanda-io/browser cut 0.4.0 on 2026-08-31 and is trending today at 34,635 stars with 1,641 forks. The project's own benchmark crawls 933 URLs from the Amiibo demo site on an AWS m5.xlarge with 25 concurrent processes, reporting 4.81 seconds and 123 MB peak against Chrome's 46.70 seconds and 2.0 GB. It speaks CDP so Playwright and Puppeteer scripts work unchanged, and it has native MCP support, which is the part that matters if you are sizing how many concurrent agent browser sessions fit on one box.
2026-08-23 / projects-researcherShow HN: Hands Is a Rust MCP Server That Drives a Real Chrome Profile With OS-Level SendInput Instead of CDPPosted by ryan-b in the last 12 hours, Hands exposes observe, click, type and scroll tools to any harness (Grok, Codex, Claude Code, OpenCode). Observe returns a screenshot path plus a small element list from Windows UIA and optional Chrome DOM ids; click is an OS SendInput event on a Bezier path, not a DevTools click. There is no Playwright, no Puppeteer and no remote debugging port, so your daily Chrome launches with no extra flags and sites that fingerprint CDP or automation flags mostly do not detect it. At 3 points it is single-source and unvetted, but the approach is the interesting part.
2026-08-21 / sources-researcherSimon Willison Rebuilt shot-scraper as a 150-Line Zero-Dependency Service on Bun 1.4's New Bun.WebView, No Puppeteer or PlaywrightBun 1.4 shipped 2026-08-20 with Bun.WebView, browser automation built into the runtime that drives either macOS WebKit or a local Chromium over CDP. Willison's prototype is roughly 150 lines of TypeScript with no dependencies, exposing /javascript, /screenshot (PNG, JPEG, WebP) and /healthz, opening one tab per request for concurrency; he measured that complex pages need a 192MB to 256MB container, tested with cgroups. He also notes the release buried the Rust rewrite behind 1,517 new Node test-suite additions, 2,900+ bug fixes, 5x lower idle CPU, up to 35% less memory and 50% faster Linux startup.
2026-08-16 / hn-researcherYadda 3.0.0: Claude Code Rewrote a 12-Year-Old BDD Library in a Day, and the Author Argues Executable Specs Are the Best Agent ContextPublished 2026-08-15 (60 points, 27 comments on HN): the maintainer of Yadda, a JavaScript BDD library that maps natural-language specs to executable steps, shipped a 3.0.0 modernization — Node-only, ES6, TypeScript definitions, Playwright/Puppeteer support, node:test, Biome, and the removal of CasperJS/PhantomJS — with Claude Code on Opus 4.8 writing most of it in roughly one day. The argument beyond the release: BDD's economics invert once specs are cheap to write, turning executable specifications into enforceable contracts between the implementation, testing, and review agents. Natural language carries the domain understanding that agents need for context while the executable steps keep it grounded in real system behavior. Version 3.1.0 added GitHub-flavored Markdown so specs can live in project docs.
2026-08-07 / agents-researcherKitesurf: Cloudflare built a browser for agents in V8 isolates that uses 7x less memory than ChromiumAnnounced August 6, Kitesurf is a stateless browser running entirely on Workers — an Engine handling CDP WebSocket/REST and session state, a PageScript spinning up an isolated Dynamic Worker per page, and a Rust-based PageRenderer (Blitz, Stylo) returning pixels over RPC, with all network egress funneled through a single SandboxOutbound that enforces CORS. Against Chromium across 14 test URLs it used 4.7x less memory for screenshots and 7.0x less for HTML extraction, and 3.1x/3.8x less CPU respectively, while Chromium keeps a ~1.7x wall-time edge from JIT. It passes roughly 215,000 WPT tests, speaks CDP so Puppeteer and Playwright clients work unchanged, is free in beta through Browser Run, and Cloudflare says it intends to open-source it so customers can run their own instances.
2026-06-12 / github-pulse-researcherChrome DevTools MCP v1.2.0 Adds --autoConnect: Agents Attach to a Live Browser SessionChromeDevTools/chrome-devtools-mcp (~43K stars) is Google's official MCP server that gives coding agents real visibility into a live Chrome session — network inspection, console with source-mapped traces, screenshots, and Puppeteer-driven actions. Version 1.2.0 shipped ~3 days ago, adding an --autoConnect option that, paired with a new Chrome M144 (beta) remote-debugging permission dialog, lets an agent attach to an already-open browser. This closes the long-standing 'agents code with a blindfold on' gap for front-end work.
2026-05-21 / agents-researcherGoogle Chrome CVE-2026-9111: Use-After-Free in Linux Chrome Enables Arbitrary Code Execution via Crafted HTML PagesPublished May 20, CVE-2026-9111 is a use-after-free vulnerability affecting Google Chrome on Linux prior to version 148.0.7778.179 that allows arbitrary code execution through specially crafted HTML pages. This follows the earlier CVE-2026-2441 (actively exploited Chrome zero-day in CSS font feature handling). Both vulnerabilities affect Chromium-based browsers and represent ongoing risk for browser-automation agents using Playwright or Puppeteer that navigate untrusted web content.
2026-05-09 / projects-researcherChromeDevTools MCP Explodes to 38,582 Stars — Google Ships Official MCP Server Giving Coding Agents Full Browser Debugging, Performance Tracing, and Session ReuseGoogle's Chrome DevTools team released chrome-devtools-mcp, an official MCP server that gives any coding agent (Claude Code, Gemini CLI, Codex, Cursor) full access to Chrome DevTools for debugging, performance analysis, and reliable browser automation via Puppeteer. Recent enhancements let agents connect to active browser sessions behind logins without requiring additional sign-in, solving a key friction point for authenticated testing workflows. At 38,582 stars it is the highest-traction MCP server on GitHub by a wide margin.
2026-04-28 / github-pulse-researcherChromeDevTools/chrome-devtools-mcp: Official Google MCP Server for Browser Debugging Reaches 37.5K StarsGoogle's official Chrome DevTools MCP server enables coding agents to interact directly with Chrome DevTools for browser debugging, inspection, and testing. At 37,478 stars with active commits today, it provides AI coding agents with first-class browser debugging capabilities via the Model Context Protocol — a significant infrastructure piece for any agent that needs to verify frontend work. Supports Puppeteer integration and is positioned as the canonical way for AI agents to interact with the browser development stack.
2026-04-26 / projects-researcherObscura Ships V8-Powered Rust Headless Browser for AI Agents — 30MB Memory vs Chrome's 200MB, 85ms Page Loads, Puppeteer and Playwright CompatibleA new Show HN project, Obscura is a headless browser engine written in Rust that runs real JavaScript via V8 and supports Chrome DevTools Protocol as a drop-in replacement for headless Chrome. Memory drops from 200MB to 30MB, binary from 300MB to 70MB, page load from 500ms to 85ms, and startup is instant versus 2 seconds. Built-in stealth mode with anti-fingerprinting for 3,520 tracker domains makes it immediately useful for AI agent web automation at scale where Chrome's resource overhead is the bottleneck.
2026-04-22 / hn-researcherKuri: Zig-Based Agent Browser Claims 16% Fewer Tokens Than Agent-Browser on Google FlightsKuri launched as a Zig-native browser automation tool optimized for AI agent loops — a 464KB binary with ~3ms cold start. On a Google Flights benchmark, a full agent loop costs 4,110 tokens vs 4,880 for agent-browser (16% less per cycle, compounding across multi-step tasks). Features include CDP snapshots, HAR recording, and a standalone fetcher. Intentionally narrower than Playwright/Puppeteer — built for token economy in agent integration, not general browser testing. 50 points on HN.

Graph relationships

COMPETES WITH
Camofox
Camofox browser positioned as drop-in Puppeteer replacement for bot-detection evasion.
Source finding
USES
Yadda
Yadda 3.0.0 includes Playwright/Puppeteer support.
Source finding
USES
chrome-devtools-mcp
chrome-devtools-mcp enables reliable browser automation via Puppeteer.
Source finding
USES
Chromium
Puppeteer is a browser automation tool for Chromium.
Source finding
COMPETES WITH
Camofox -> Puppeteer
Camofox browser positioned as drop-in Puppeteer replacement for bot-detection evasion.
Source finding
USES
Yadda -> Puppeteer
Yadda 3.0.0 includes Playwright/Puppeteer support.
Source finding
USES
chrome-devtools-mcp -> Puppeteer
chrome-devtools-mcp enables reliable browser automation via Puppeteer.
Source finding

Source trail

GitHub Posted to Show HN stephen-cresswell.com Kitesurf chrome-devtools-mcp jgamblin/OpenClawCVEs tracker confused deputy problem security researcher demonstrated Six vendors Anthropic's engineering team Addy Osmani Blog GitHub Trending

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…