Mozilla's developer arm launched Tabstack this week, pitching it as "the web execution layer for AI agents." The API handles browser rendering, navigation, and data extraction so developers can skip the headless Chrome headaches.
Three endpoints do the heavy lifting: /extract converts pages to Markdown or structured JSON, /generate transforms content on the fly, and /automate lets agents click, scroll, and fill forms. The TypeScript SDK is already on GitHub. Under the hood, Tabstack routes requests dynamically, starting with lightweight fetches and only spinning up full browser sessions when JavaScript rendering demands it.
Pricing runs $1 per 1,000 Markdown extractions, $5 for JSON, and $7.50 for automation actions (which can chain multiple steps per call). The free tier includes 50,000 credits monthly. That's roughly 5,000 pages in Markdown or 500 automation tasks, per Mozilla's own math. No word yet on the CAPTCHA-solving feature mentioned in early coverage, which appears limited to paid plans.
Mozilla is leaning hard on privacy as the differentiator. Data is "ephemeral by default," requests use a dedicated User-Agent for site owners to track, and robots.txt opt-outs are honored. Most notably: Mozilla says it won't train models on scraped content.
The Bottom Line: Tabstack enters a crowded field (Browser Use, Browserbase, AgentQL) but Mozilla's privacy stance and nonprofit backing could matter to enterprise buyers wary of data handling.
QUICK FACTS
- Launched: January 14, 2026
- Free tier: 50,000 credits/month
- Extract pricing: $1 per 1k (Markdown), $5 per 1k (JSON)
- Automate pricing: $7.50 per 1k actions
- SDKs: TypeScript and Python available
- Discord community: 5,400+ members
