The End of Captchas – Agents Control Firefox via API Like a Human

Mozilla has announced the TABS API for Firefox, through which agents can control the browser: clicking through pages, extracting content and converting it to JSON or Markdown, answering questions, and filling out forms. Developers can design agentic workflows or simulate human page visits for testing purposes.

The API performs tasks that developers specify in natural language. According to the website, no separate LLM is required for this, although the service is subject to a fee beyond a small free volume. A waiting list is currently open for interested parties, and Mozilla has not yet published prices; handling of captchas and proxies is only available in the paid Pay-as-you-go package.

Developers can access four endpoints in the TABS API:

• /Extract (POST): Delivers the content of web pages in JSON, Markdown, or another desired format
• /Generate (GET): Transforms content, for example from a blog entry to a social media post
• /Automate (GET): Complex, interactive, and agent-controlled tasks
• /Research (POST): Context-oriented web search with an answer

Mozilla emphasizes a strong focus on data minimization and security. The API transmits only the most necessary data, encrypted end-to-end via TLS. All data no longer needed is deleted immediately.

SDKs for TypeScript and Python

Developers can access the TABS via Curl or use an SDK provided by Mozilla for TypeScript or Python. An example for TypeScript provides the documentation:

import { Tabstack } from '@tabstack/sdk';

const tabs = new Tabstack({
  apiKey: process.env.TABSTACK_API_KEY!
});

// Access generate methods
tabs.generate.json(url, schema, instructions, options);

(who)