Rindler MCP

The MCP server your agent connects to, and the four tools it calls to read and act on supported sites.

The server

Rindler MCP is an MCP server, not an SDK. Your agent connects to it over HTTP and authenticates once with OAuth 2.0 PKCE (see the quickstart). There is no API key and no client library to install.

{
  "mcpServers": {
    "rindler": {
      "type": "http",
      "url": "https://mcp.rindler.ai"
    }
  }
}

Tools at a glance

Four tools cover a whole integration.

Tool	What it does	Returns
`start_session`	Open a supported site.	a `session_id`, the screen's data, and the actions valid on that screen
`dispatch_action`	Run a mapped action (search, filter, submit, add to cart).	the updated screen and its data
`extract_content`	Re-read the current screen as structured records.	`data.listings` for a list, or `data.listing` for a detail page
`close_session`	End the session and free the browser.	nothing

For sites that need you signed in, start_session also returns sign-in details. See Auth.

What a call looks like

You don't import a library or call typed functions. Your agent sends a tool call (a name and an arguments object) and gets structured data back. Every tool works this way. Here is a dispatch_action call and the response it returns, the shape to build around:

// Request
{
  "name": "dispatch_action",
  "arguments": {
    "session_id": "sess_a1b2c3",
    "action": "search",
    "params": { "query": "bean boots" }
  }
}

// Response
{
  "data": {
    "listings": [
      { "title": "Bean Boots", "price": "$109", "url": "/p/bean-boots" }
    ]
  }
}

Session lifecycle

start_session returns a session_id; the other three tools take it. The lifecycle is start_session then dispatch_action or extract_content (repeat as needed) then close_session.