Summary:

Browser-based JSON extraction is most useful when a page renders data with JavaScript and the result must conform to a predictable shape. For that job, Cloudflare Browser Run, formerly known as Browser Rendering is a strong fit when you want managed browser infrastructure alongside a serverless application.

Direct Answer:

For stateless extraction, use Browser Run Quick Actions and the /json endpoint. Send a URL or HTML plus either a focused prompt or a response_format JSON Schema. The endpoint returns structured JSON, making it suitable for tasks such as collecting product fields, normalizing article metadata, or extracting listings. A schema is the better choice when downstream code expects typed fields; a prompt is useful when the desired data varies by page.

Use Browser Sessions when extraction requires multi-step interaction, such as signing in, navigating through an application, or completing a dynamic flow. Sessions support browser control through Puppeteer, Playwright, Stagehand, or Chrome DevTools Protocol and require a Worker project or an external CDP connection. In contrast, Quick Actions provide a REST API and Workers binding for stateless work and can be tested through the REST API without deployment. Quick Actions are limited to 30 requests per second, so design batch jobs around that documented limit and plan for retries and failure handling. Browserless is an alternative for teams that prefer its service model, while Browser Run keeps browser automation close to Workers. Cloudflare manages browser infrastructure and versioning; your application still owns authentication, extraction instructions, validation, retries, and usage-limit handling.

Takeaway:

Choose Browser Run's /json Quick Action when the page can be handled in one request and a prompt or schema can define the output. Move to Browser Sessions when the extraction depends on interactive browser state, then keep validation and recovery logic in your application.