A real browser, as a step in your workflow
Not an HTTP fetch — a real browser that runs the page's JavaScript and waits for it to finish, then hands back what a person would actually see. It is a component you drop onto the canvas, its arguments can be drawn from earlier steps, and your plan's allowance is checked before the request leaves rather than discovered on the invoice.
A real browser as a canvas component — runs JavaScript, returns rendered content or a screenshot, with a per-workspace allowance checked before each request.
What the component gives you
- A real browser, not a fetch — Plenty of pages are empty until their JavaScript runs. This is a real browser that loads the page, runs its scripts and waits for the content to settle, so you read what a person would see rather than the bare HTML a plain HTTP request returns.
- Rendered content or a screenshot — Get the page back as clean markdown ready to feed the next step or an agent, or capture a screenshot of the finished render. You choose how the page comes back, per step, with no separate parsing stage in between.
- Arguments that read your data — The URL to open is rarely a constant. Each argument can be an expression over the payload or an earlier step's output, so a page opened from a record you just looked up does not need a transformation step in front of it purely to rename a field.
- Usable by a workflow or an agent — The same component is available as an ordinary step you place in order, and as a tool you can grant to an agent that decides for itself when a page is worth opening. Both routes are metered the same way, so an agent cannot spend past the ceiling a workflow respects.
Platform capabilities
- Real browser: How each page is loaded
- Step or tool: Two ways to reach it
- Before the call: When your allowance is checked
Do I have to host anything?
No. The browser runs as part of the platform, so there is no container to deploy, no browser to keep patched and no separate bill. You place the component and configure what it should open.
Does it run the page's JavaScript?
Yes. It is a real browser, so scripts run and the page is given time to finish loading before anything is read. You can take the result as markdown or as a screenshot of the finished render, and pages that are blank without JavaScript come back filled in.
Can an AI agent decide what to open?
Yes, if you grant it. Connect the component to an agent and the model may call it when it judges a page worth reading. It is metered exactly as a workflow step is, so granting the capability does not raise the ceiling — and if you would rather decide yourself, use it as an ordinary step instead.