New Test
Preview
Content preview will appear here as you type…
Previous Tests
0
Internal QA toolkit powered by GPT-4o mini
Enter a Freshdesk ticket number or URL. The orchestrator fetches the ticket, identifies the worked-on site, runs technology detection, and performs focused Visual QA or a Lighthouse-style speed audit when the ticket is performance-related.
Paste a note describing the frontend change to verify, and the page URL. Runs Playwright Chromium screenshots + visual QA only — no Freshdesk ticket fetch.
Enter a Freshdesk ticket for WordPress/WooCommerce backend QA. Verifies WP admin settings, plugin configuration, logs, and related backend work — no public visual QA.
Automate checkout process on HTTP Basic Auth protected sites. Uses Arizona test data automatically.
Classic path — agents drive Chromium with Playwright directly. Kept for comparison and future use alongside the Browser Gateway protocol.
Browser Gateway automation — a general planner reads your instructions,
observes the live page (fields + links), then builds and runs the needed steps
(click, type, execute_js, screenshots).
Include the starting website URL inside the instructions.
Extension WebSocket: Loading…
If no Chrome agents are connected (or they stop responding), the server
may run the test with Playwright backup (headless Chromium on Railway).
Set PLAYWRIGHT_FALLBACK_IF_NO_CHROME=true on the server to enable this.
Standalone browser-use LLM agent — it drives its own local Chromium browser directly, with no Chrome extension / Browser Gateway required. Describe the task in plain language, including the starting website URL.
Server outbound IP: checking… — headless/Railway runs use this IP; if a target site 403s only there (not via a CDP Worker), allowlist it. Not static on Railway Hobby — can change on redeploy.
Runs headless on Railway by default (fast, accurate, no CDP Worker needed) — it only reaches for a registered QA Station if the site looks blocked or inaccessible. To instead always drive a specific real browser, uncheck "Run without CDP Worker" below and pick one, or click "Download the CDP Worker package" above to turn any machine (yours, a teammate's, anyone's) into one in a couple of minutes — a ZIP with everything needed, pre-filled with this server's address; you can also just hand it to an AI coding assistant and ask it to set it up.
Any model id your OpenAI key has access to works — this list is just a shortcut, not a restriction.
Runs a browser test task on a connected QA Station machine.
Headless on Railway is the default — fast, accurate, and not dependent on a station being online. It only escalates to a registered QA Station if the site looks blocked or inaccessible. Pick a specific station below only if you want this run to always use a real desktop browser.
Paste a ticket link. The pipeline reads the ticket, finds the website and environment from the latest agent note, applies the standing QA instructions for the ticket's Type, and drafts a structured QA prompt. Nothing runs until you review the prompt and press Run QA.
Paste a rough QA instruction — the way a support agent would write it — and get back a structured prompt (OBJECTIVE / TARGET URL / STEPS / EXPECTED RESULT / PASS & FAIL CRITERIA / EVIDENCE) that the browser agent executes reliably. No ticket needed. Nothing runs from here; copy the result or send it to the AI Browser Agent.
OpenRouter entries only work while AI_PROVIDER=openrouter is set on Railway — otherwise pick an OpenAI model above.
LangGraph pipelines, sub-agents, and shared tools. Hover any node for details.
For an interactive n8n-style workflow map with demo animation, open
System Map.
Update frontend/agents-registry.js when adding new agents.
Reusable utilities and state schemas
Server and command-line entrypoints