AIToday
Large Language ModelsAI Coding AssistantsAI Business & IndustryZenn AI/MLPublished: Oct 16, 2026, 22:01 JST

Claude Code drives no-API low-code console via Chrome extension

Claude Code drives no-API low-code console via Chrome extension

3 Key Points

  1. What happened

    An e-dash SRE used Claude Code with a Chrome extension for about three months on a low-code admin console with no connector and no API, delegating action creation, saving, test runs and result counts. The human clicked only once, to confirm the save dialog.

  2. Why it matters

    Anything reachable only through a logged-in browser screen — not through an MCP connector or an API — becomes automatable again, though the person keeps responsibility for staging and production.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, in your inbox every morning.

Context & Analysis

Automation keeps running into small in-house tools that were picked to fit a workflow and never got an API. For a low-code console, an action is defined only through its screen — there is no API to create one — so the work stayed with a person opening the page.

The author's approach was to let Claude read the ordinary, already-logged-in Chrome rather than sign in again. Read access to console errors and network traffic let it distinguish a page that merely displays from one whose call returns a 500, and the same round trip covered saving, a test run and a result count.

Just as specific is what remained manual. Credentials never enter Claude Code's context, and it stops at login screens and CAPTCHAs. All of the described work sits in development environments; staging and production stay human. Dialogs and cross-origin iframes are named as reach limits, and fixed skills divide the steps into creating, fixing, checking on screen and checking data.

FAQ
What does Claude Code actually do in the browser?
It reads displayed elements and text, takes screenshots, clicks, types, navigates and opens or closes tabs. It also reads console errors and network requests and can record a flow as a GIF.
What can't Claude Code do in this setup?
While a browser or OS confirmation dialog is open, screenshots, coordinate clicks and page navigation all fail. Text inside cross-origin iframes also cannot be read, so result correctness there is checked by screenshot.
How do you turn the Chrome integration on?
Start Claude Code with --chrome, or run /chrome and choose Enabled by default. In the VS Code extension, typing @browser in the prompt also works.
Why not use Claude Code's built-in browser?
The built-in browser is isolated from everyday Chrome, so it cannot reach logged-in services. Since the target is an internal admin console, the extension is used instead.

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.

Questions and answers are published on this page.

Related Articles

Next articlePerplexity to open-source SPACE agent sandbox