Browserbase Changelog

Use the updated Fetch API to fetch and return web content as markdown or structured JSON from any URL. Faster and cheaper than spinning up a browser, in formats that are easier for humans and models to read.

Data cap raised from 1MB to 5MB.

Extract ~$4 / 1k pages, ~$7 / 1k with proxies.

Session replays can now be embedded directly in your product and streamed to your end users within seconds of a session ending.

Includes:

• API for embedding session replays in your own product
• Playback in any HLS-compatible player
• Storage, fMP4 encoding, and CDN delivery handled by Browserbase

Your backend fetches the session's .m3u8 playlist from our API, your frontend hands it to a player, and end users stream segments directly from our CDN.

Free on every plan, up to 120 sessions per minute.

Stagehand 3.4.0 is live.

Use ignoreSelectors to keep noisy parts of a page out of extract() and observe(): ads, nav, modals, related posts, and anything else your agent should ignore.

Plus:

• Agent variables are now supported in the Stagehand API schema without the experimental requirement
• Agent mode now defaults to hybrid for compatible models and DOM mode otherwise
• New CUA model support: openai/gpt-5.4-mini, openai/gpt-5.5, and anthropic/claude-haiku-4-5
• Better OOPIF frame handling and stronger observe element ID prompting

Every file your agent downloads is now individually addressable.

The Downloads API now treats every file as a first-class object with its own ID and metadata. You can list downloads for a session and filter by filename, MIME type, size, or timestamp; fetch an individual file or just its metadata; and delete a single file on demand.

The session-level zip endpoint continues to work, no migration required. Your existing downloads are already addressable.

Stagehand 3.3.0 is live.

Verified mode for improved agent identity. Sessions get a verifiable identity that helps on sites that gate bot traffic.

Plus:

• Adaptive thinking with Anthropic models. Claude CUA now scales the extended-thinking budget to match task complexity — faster on simple steps, deeper when it counts
• Agent usage in stagehand.metrics. agent.execute() calls now contribute to stagehand.metrics on API-backed sessions, so token counts and model calls land where you expect them
• Strict structured outputs. Supported model paths now use strict JSON schema enforcement, eliminating malformed responses on extract() and other structured calls
• Clearer file upload elements. File inputs are now explicit in the page snapshot, so agents can identify and drive upload flows reliably

Managing model providers adds friction to every Stagehand setup. Model Gateway removes that layer by handling routing, retries, and billing at market price with no markup. Pass your Browserbase API key and pick a model—switch between OpenAI, Anthropic, and Gemini with a one-line change. No new accounts, reworked setup, or extra friction needed.

Browserbase and Prime Intellect partnered to make browser agent training more accessible. BrowserEnv, a reinforcement learning environment built on Browserbase's infrastructure, is now available at browserenv.com, providing scalable browser environments and live website access for training.

Browserbase Search, powered by Exa, enables agents to search the web with a single API call. Send a plain-text query and get back URLs your agent needs to act on next. Results return URL, title, and description (up to 25 results). Each plan includes 1,000 free searches.

Free users can now use up to 3 browsers concurrently, increased from the previous limit of 1 browser.

The Browserbase MCP server was migrated to Browserbase-managed infrastructure to improve reliability and simplify updates. MCP is now served from Browserbase-managed endpoints with improved stability for longer-running sessions. No action required for existing users.

Introducing the Fetch API for Browserbase. Provide a URL and receive the page content without requiring a browser session. Fast, cheap (~$1 per 1,000 pages), and designed for agents that need to read the web efficiently.

Stagehand now automatically caches results of repeated actions to eliminate redundant LLM calls. When encountering a page structure it's seen before, it serves the cached result instantly. Results in up to 2x faster execution and ~30% cost reduction on repeat workflows. Caching is conservative by default and scoped to your project.

Browserbase is now a one-click integration on the Vercel Agent Marketplace. Agents deployed on Vercel Sandbox can connect to remote browsers over CDP. Also launching Web Bot Auth support with Vercel, enabling agents to send cryptographically signed requests that Vercel verifies, allowing reliable browsing of Vercel-hosted deployments without security blocks.

Deploy agents and automations directly to Browserbase to run alongside their browser sessions. Define, deploy, and invoke functions while Browserbase handles execution, browser lifecycle, and results. Benefits include up to 70% faster latency, fewer moving parts, async/long-running executions with results tied to real browser sessions, and local development mirroring production. Available in TypeScript.

Session recordings rebuilt from the ground up using CDP-based screencast instead of DOM-based replay. Now offers pixel-accurate screen recordings, event-driven frame capture, asynchronous constant-time video encoding with HLS, seamless multi-tab playback, secure signed segment delivery with CDN enforcement, and just-in-time encoding for low costs and fast playback.

Stagehand v3 extends support across all languages including Python, Go, Java, Ruby, and Rust. The canonical implementation of Stagehand powered by the newest features benefits every SDK immediately, enabling faster shipping and improved extensibility. Includes revamped Python with v3 support.

Major release featuring a new foundation for Stagehand with improved speed and flexibility. Includes 20-40% speed increase across act, extract, and observe calls; compatibility with Playwright, Puppeteer, and Patchright; automatic action caching; non-AI primitives (page, locator, frameLocator, deepLocator); bun compatibility; simplified extract schemas; CSS selector support; targeted extract and observe across iframes and shadow roots; and more intuitive type names.