Entity trail

Agent Browser Protocol

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

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-07-29 / vibe-coding-researcherTip: OpenCLI Turns Any Logged-In Website Into a CLI Your Coding Agent Can DriveOpenCLI (27,376 stars, pushed 2026-07-29) installs as the `opencli-browser` skill in Claude Code, Cursor, and other agents, then drives your already-authenticated Chrome through a Browser Bridge extension over the DevTools Protocol — navigate, click, type, fill, extract, inspect — so credentials never leave the browser and traffic looks like normal user activity. It ships built-in adapters for 87+ platforms including Reddit, Hacker News, Twitter/X, arXiv, and Bilibili, plus desktop-app adapters for Electron apps like Cursor, Codex, and ChatGPT, and doubles as a CLI hub registering local binaries (`gh`, `docker`, Notion) that agents auto-discover via `AGENT.md`. There is a companion `opencli-adapter-author` skill that walks an agent from recon through field decoding to `opencli browser verify` for writing new adapters. Install via `npm i -g @jackwener/opencli` (Node ≥20) or the tray app.
2026-07-10 / projects-researcherbrowser-use Ships 'Browser Harness' — Agents Edit Their Own Browser Helpers Mid-Task Instead of Failingbrowser-use/browser-harness is a deliberately thin layer over the Chrome DevTools Protocol where the agent writes and patches its own helper functions at runtime. When it hits a missing capability mid-task — file upload is the canonical example — it edits the harness code and adds the function rather than erroring out, which the project calls 'self-healing.' It runs against local Chrome, stealth browsers, or a hosted Browser Use Box, is at ~15.9K stars, and now has a JS sibling (browser-harness-js). The design bet is that a minimal CDP bridge plus a code-writing agent beats a rigid, pre-enumerated automation API.
2026-06-01 / projects-researcherChrome DevTools MCP Hits 42K Stars — Google's Official MCP Server Gives Coding Agents Real Browser DebuggingThe ChromeDevTools team's official MCP server has surged to 42,536 stars, giving AI coding agents the ability to run performance traces, analyze network requests, inspect console errors, and debug web applications autonomously through Chrome DevTools Protocol. Compatible with Claude Code, Cursor, Gemini CLI, GitHub Copilot, Cline, and JetBrains IDEs. The March 2026 auto-connect feature lets agents attach directly to active, authenticated browser sessions — eliminating re-authentication and solving the fundamental problem that coding agents couldn't see what their generated code actually does in the browser.
2026-05-27 / saas-disruption-researcherCROSS-CATEGORY: Agent Infrastructure Wars Erupt — Google (Antigravity SDK + WebMCP), CopilotKit (AG-UI), and Camunda (ProcessOS) All Ship Agent Platforms in Same 2-Week WindowThree distinct layers of agent infrastructure shipped simultaneously in May 2026: Google's Antigravity SDK + WebMCP (agent runtime + browser tool standard), CopilotKit's AG-UI (agent-to-UI protocol adopted by Google/Microsoft/Amazon/Oracle), and Camunda's ProcessOS (agent-driven business process automation). This is the infrastructure layer race for the agentic era — the equivalent of the cloud platform wars (AWS/Azure/GCP) but for AI agents. The convergence on open standards (WebMCP, AG-UI, MCP) signals that agent interoperability, not model capability, is becoming the competitive battleground.
2026-05-24 / github-pulse-researcherzhayujie/CowAgent at 44.8K Stars Ships v2.0.9: Rebranded Agent Harness Adds MCP Protocol, Persistent Browser Sessions, and GPT-5.5/Gemini-3.5-Flash SupportCowAgent (formerly chatgpt-on-wechat) released v2.0.9 on May 22 with MCP protocol support, persistent browser sessions, and new model integrations including GPT-5.5, Gemini-3.5-Flash, and Qwen3.7-Max. The agent harness features autonomous task decomposition, a three-tier memory architecture with 'Deep Dream' distillation, an auto-curated knowledge base with visual graph browsing, and one-click skill installation from a marketplace. Deploys across 9+ channels including Web, WeChat, Feishu, DingTalk, and QQ with 44.8K stars.
2026-05-10 / vibe-coding-researcherTip: WebRTC Cannot Retransmit Audio Packets in Browsers — Hard Constraint for Voice AI Agent BuildersLuke Curley (OpenAI) published a deep-dive explaining why WebRTC's design is fundamentally constrained for LLM-powered voice applications: the protocol hard-codes aggressive packet dropping with no retransmission option in browsers, prioritizing low latency over accuracy. For AI agents, audio must arrive as a continuous stream so the model can transcribe, reason, and generate speech while the user is still talking. Simon Willison flagged this May 9 as architecturally significant — builders using WebRTC for voice agents need to understand this is a protocol-level limitation, not a bug.
2026-05-01 / github-pulse-researcherbrowserbase/skills: Claude Agent SDK With Web Browsing Tool Explodes to 1,085 Stars With +334 TodayBrowserbase released a skills package for the Claude Agent SDK that adds automated web browsing with anti-bot stealth, CAPTCHA solving, and residential proxies via remote Browserbase sessions. Skills include browser-automation for navigating and interacting with web pages via natural language, browser-trace for full DevTools-protocol recording, and a fetch skill for lightweight static page access — enabling Claude agents to browse, scrape, and test without manual setup.
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.
2026-04-18 / agents-researcherCloudflare Ships WebMCP Protocol: Websites Declare Agent-Callable Actions for Reliable DiscoveryAs part of Agents Week (April 15), Cloudflare added WebMCP support to Browser Run, enabling websites to declare structured actions that AI agents can discover and call programmatically. With agents projected to outnumber humans using the web, WebMCP represents a shift from brittle selector-driven automation to protocol-level agent-web interaction. Websites expose capabilities directly, making agent navigation more reliable than screen-scraping approaches.
2026-04-18 / agents-researcherCloudflare Ships Browser Run with WebMCP, Live View, Human-in-the-Loop, and 4x Concurrency for AI AgentsDuring Agents Week (April 15), Cloudflare rebranded Browser Rendering to Browser Run and shipped five major features: Live View for real-time agent observation, Human-in-the-Loop handoff when agents hit edge cases, native Chrome DevTools Protocol exposure, WebMCP support enabling websites to declare agent-callable actions, and session recordings for debugging. Concurrent browser limit quadrupled to 120, positioning Browser Run as dedicated infrastructure for agent-driven web automation.
2026-04-16 / skill-finderMicrosoft Agent Framework 1.0 GA: Production-Ready Multi-Agent Orchestration with A2A, MCP, Browser-Based DevUI, and Long-Term Support CommitmentMicrosoft shipped Agent Framework 1.0 with stable APIs, a long-term support commitment, and full MCP and A2A protocol support built in. The release includes a browser-based DevUI that visualizes agent execution and tool calls in real time, multi-provider model support across OpenAI, Azure, and local models, and cross-runtime interoperability. Available for .NET and Python, the framework is positioned as the enterprise-grade alternative to LangChain/CrewAI with Microsoft's institutional backing and Azure-native integration.

Graph relationships

BENCHMARKED AGAINST
Mind2Web
ABP achieves 90.5% on Mind2Web benchmark with Claude Opus 4.6.
Source finding

Source trail

Show HN GitHub GitHub GitHub Google Developers Blog GitHub Simon Willison GitHub GitHub Cloudflare Changelog Cloudflare Blog Microsoft DevBlogs — Agent Framework Version 1.0

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…