Automate browser tasks by reading, clicking and filling forms

The poster uses Claude in Chrome to read page content, click buttons and fill in forms, and asks about alternatives for browser automation.

What They Did

The poster says they use Claude through a Chrome extension to automate browser work by reading page content, clicking buttons, and filling forms, then asks what other setups people use for the same workflow. In the replies, people point them to alternatives like chrome-devtools MCP, Browser Use, Stagehand, Playwright, Browserbase-style managed browsers, and Anthropic’s computer-use API, with several commenters stressing screenshots, logs, retries, and human approval for risky steps.

How It Went

There is no reported result, metric, or finished project in the thread; it is mainly a request for alternatives and a comparison of tools. The clearest limitations mentioned are flaky behavior, no multi-tab parallel work in one MCP setup, bot detection, slowness, and cost.

Worth Knowing

The most practical catch commenters repeat is reliability: use human approval for money or account actions, and confirm page state with screenshots before the agent acts.