n8n-nodes-testmuai

Verified n8n community node for TestMu AI Browser Cloud (formerly LambdaTest). Lets your AI Agent drive real browsers — Chrome, Firefox, Safari, Edge — across Windows, macOS, and Linux. Every session runs live and can be viewed on the Browser Cloud dashboard with video replay, console logs, and network capture for full debugging visibility.

What's in the package

One node: TestMu AI (Formerly LambdaTest) Agent — a tool that any AI Agent calls to drive a real cloud browser. Connects to TestMu's cloud via the W3C WebDriver protocol over HTTPS, so it works inside n8n Cloud's verified-node sandbox.

Looking for the script-runner or in-process Playwright agent? Those features live in the companion package n8n-nodes-browsercloud (self-hosted n8n only).

Install

In your n8n instance:

• Self-hosted: Settings → Community Nodes → Install → enter n8n-nodes-testmuai
• n8n Cloud: Available in the Verified Community Nodes marketplace.

Credentials

Add a TestMu AI (Formerly LambdaTest) API credential with:

• Username — your TestMu AI username
• Access Key — your TestMu AI access key

Find both in your TestMu account profile.

How to use with an AI Agent

[Trigger] → [AI Agent]
              ↑ Tools (sub-input on the bottom of the Agent node)
              [TestMu AI (Formerly LambdaTest) Agent]
              ↑ Chat Model
              [Gemini / Claude / OpenAI Chat Model]

1. Add an AI Agent node.
2. Wire any tool-calling-capable Chat Model (Gemini 2.5 Flash, GPT-4o, Claude Sonnet) into the Agent's Chat Model socket.
3. Wire TestMu AI (Formerly LambdaTest) Agent into the Agent's Tools socket.
4. Pick your TestMu credential (available on your TestMu AI dashboard) on the tool node.
5. Configure browser, platform, region (US/EU), and version.
6. Give the Agent a goal in its user message — e.g. "Open news.ycombinator.com and tell me the title of the top story."

Suggested system message for the AI Agent

You drive a real cloud browser via the TestMu AI Agent tool. Steps:

1. Call navigate first to open the right URL.
2. Click and type refer to elements by their ref number from the latest snapshot.
   Every tool response includes a fresh snapshot — use it.
3. If an element's role contains "(readonly)", do NOT try to type into it.
   Click it instead — it usually opens a picker with a real input you can type into.
4. Use get_text to extract content.
5. When the goal is achieved, ALWAYS call action=release as your final tool call.
   This frees the cloud browser immediately instead of letting it idle-time out. You can also type "Release session after completion" in natural language to achieve the same.

Never refer to refs from previous turns — refs are only valid against the latest snapshot.

Actions exposed

| Action | Purpose | |---|---| | navigate | Open a URL | | snapshot | Return a numbered list of interactive elements on the current page | | click | Click an element by its ref number | | type | Type text into an input by ref (optionally pressing Enter after) | | get_text | Read text from a ref or the whole page | | screenshot | Capture a base64 PNG of the current viewport (useful for vision models) | | release | End the cloud browser session. The AI Agent should call this when the goal is achieved. |

After every action, the response includes a fresh page snapshot — so the AI Agent rarely needs to call snapshot separately between every click and type.

Architecture notes

• Pure HTTP — no third-party Node dependencies, no subprocess spawning, no in-process Playwright. All browser interaction goes through TestMu's WebDriver Hub via n8n's sandboxed httpRequest helper.
• Session state lives in n8n's workflow static data, scoped by execution ID. Stale sessions auto-prune after 30 minutes.
• Elements are tagged with data-n8n-ref attributes by the snapshot script so click / type can find them again reliably via a single CSS selector lookup.
• Live session videos and console / network logs are visible on the TestMu AI's Browsercloud dashboard.

Limitations vs the companion n8n-nodes-browsercloud package

• No script-runner (cannot exist in the verified sandbox — would need child_process and fs).
• No in-process Playwright (network roundtrip per action vs in-memory call).
• Full-page screenshots return viewport-only (W3C WebDriver doesn't standardize full-page; we report fullPageRequested: true so callers know what was asked for).

For workflows that need any of the above, install the companion n8n-nodes-browsercloud package on self-hosted n8n.

License

MIT