AutoQAI

Then register it with your AI CLI agent (Claude Code, Cursor, etc.). The agent talks to the MCP server, the MCP server talks to this extension over a local-only (loopback 127.0.0.1) WebSocket, and the extension drives your tab. Without the MCP server installed and paired, the extension has nothing to connect to. 🔌

👉 Read the full setup guide, supported agents, and docs here: https://utilsware.com/products/autoqai

────────────────────────────────── 👥 WHO IT'S FOR ────────────────────────────────── 👨‍💻 Developers running Claude Code (or any MCP-capable agent) who want the agent to reproduce and fix bugs inside their own browser session. 🧑‍🔬 QA engineers who want live console/network observability, screenshot-based bug capture, and a one-click "ask the agent to fix this" loop. 🧱 Anyone who has hit the wall of "the AI's test browser is not logged in to my staging environment".

────────────────────────────────── ✨ WHAT YOUR AGENT CAN DO ────────────────────────────────── 🖱️ Interact with the page — click, fill, type, press keys, select options, hover, drag, and upload files. 📸 Capture screenshots of the page or specific elements/regions (PNG or JPEG). 🌳 Read accessibility/DOM snapshots so the agent understands page structure without guessing. 🧪 Evaluate JavaScript in the real page context. 🧾 Extract and compare tables, forms, charts, sections, and styles — old vs new, before vs after, tab vs tab. ✅ Run assertions — text, values, visibility, element counts, classes, styles, enabled/checked state, network success, and "no console errors". 🐞 Read console logs, network activity, and page issues to diagnose real bugs. ⏳ Wait intelligently — for selectors, network idle, or custom JS conditions — so tests are stable, not flaky. 🗂️ Work across multiple tabs and windows in parallel, each bound to a project with a short human-readable label (e.g. main, qa-tenant, mobile). 📋 Manage a queue of testing tasks and bug reports, with run history and findings saved per project. 🎯 Element pick + region capture, packaged into a structured bug payload. 🎫 Browse Jira tickets read-only through your own authenticated session.

────────────────────────────────── 🧩 HOW IT WORKS ────────────────────────────────── 1️⃣ Install the MCP server: npm install -g @autoqai/mcp 2️⃣ Install this extension and pin it. 3️⃣ In your project directory, start the AutoQAI MCP server — it prints a port and a pairing token. 4️⃣ Open the extension popup, paste host/port/token, pair & connect, then bind the tab you want tested. 5️⃣ In your AI CLI agent, AutoQAI appears as a regular MCP server. Ask it to test, reproduce a bug, or compare two screens — it drives your real tab and reports findings back. The agent can target a specific tab by its label.

────────────────────────────────── 🛡️ PRIVACY & SECURITY ────────────────────────────────── 🏠 All communication is local-only. The extension only accepts connections from 127.0.0.1 (loopback); non-local connections are refused outright — no LAN exposure. 🔑 Pairing uses a per-project token generated by the MCP server and entered once in the extension. 🪪 The Chrome debugger attaches only after you bind a tab, and Chrome's standard "is being debugged" banner stays visible the whole time. 🚫 No credentials, OAuth tokens, or session cookies are stored or transmitted by the extension. 📡 No analytics, no telemetry, no third-party SDKs. Page content (DOM, screenshots, console logs) only leaves your browser when the AI agent you started explicitly calls a tool that returns it. 👤 The agent acts inside your existing authenticated session — it never launches a separate browser or asks for your passwords.

📖 Full documentation, supported agents, and setup instructions: https://utilsware.com/products/autoqai/docs