Docs

Let your agent use real websites.

Rindler turns supported sites into clean tools your agent calls to search, read prices, fill forms, and check out. No scraping, no selectors, no API key.

Start here

Quick Start

Connect your agent and run your first task on a supported site.

Rindler MCP

The four tools your agent calls to read and act on a site.

Mapping new sites

Need a site Rindler doesn't support yet? It can be mapped on demand.

Core tools

Four tools cover a whole integration. Open a session, act on the page, read structured data back, and close it.

start_session

Open a supported site. Returns its data, a session_id, and the actions valid on that screen.

dispatch_action

Act on the screen (search, filter, submit, add to cart). Returns the updated data.

extract_content

Re-read the current screen as structured records.

close_session

Release the browser when you're done.

dispatch_action and extract_content return structured records inline, so you rarely touch raw HTML: a response is a few hundred tokens of JSON (data.listings for a list, data.listing for a detail page), not the 50k-token DOM a raw fetch returns. start_session returns the screen's state plus the session_id you pass into every later call. Auth, navigation, retries, and error recovery run server-side.

What these docs cover

Connecting an agent, reading data, and taking actions on supported sites, plus mapping a new site and signing in to gated sites. See the current catalog at rindler.ai.

SLAs and quotas are not published here. For anything else, reach founders@rindler.ai.