See what APIs a page calls —
recorded as you click.

What you get

Every recording writes three files. The one you read is the Markdown doc.

What the doc looks like

A slice of a real generated .md — overview, diagram, ordered call table, per-call detail.

{
  "cart_id": "c_8123",
  "email": "you@example.com",
  "card_number": "[redacted]",
  "cvv": "[redacted]"
}

Install in 10 seconds

No Node.js, no PATH headaches. One command installs a standalone binary — on macOS, Windows or Linux.

macOS & Linux — Terminal:

curl -fsSL https://apiflowtracker.com/install.sh | sh

Windows — PowerShell:

irm https://apiflowtracker.com/install.ps1 | iex

The installer prints the exact MCP config to add to Claude/Cursor — it looks like this (macOS/Linux path shown):

{
  "mcpServers": {
    "browser-flow-tracker": {
      "command": "/usr/local/bin/browser-flow-tracker"
    }
  }
}

Already have Node? You can instead use "command": "npx", "args": ["-y", "browser-flow-tracker@latest"].

1. Install & add the config above, then restart your AI app.
2. Say the phrase: “let's record the session for this url …”. A real browser window opens.
3. You navigate the flow — log in if needed (remembered next time).
4. Say “done” or close the browser. The doc + JSON + HAR are written for you.

Why it's better than the console

It speaks the browser's own DevTools Protocol, then does the tedious parts for you.