HAR capture

WebSummoner can record everything a session requests and write it as a HAR 1.2 file, readable by browser devtools, Charles, Fiddler and most reporting tools.

Start the hub with a directory to write to:

websummoner -har-output-dir /opt/websummoner/har

then ask for it per session:

{
  "capabilities": {
    "alwaysMatch": {
      "browserName": "chrome",
      "websummoner:options": {
        "enableHAR": true,
        "harName": "checkout-flow.har"
      }
    }
  }
}

harName is optional; without it the file is named after the session id.

The file is written when the session ends, and served from the hub:

Request	Result
GET /har/<name>.har	The file
GET /har/?json	A JSON listing of every HAR file
DELETE /har/<name>.har	Removes it

How it works

The hub subscribes to the browser’s own CDP Network domain for the life of the session — there is no proxy in front of the browser, so HTTPS needs no certificate and nothing about the page’s network behaviour changes.

This makes HAR capture Chromium-only — Firefox and WebKit expose no CDP endpoint, and requesting enableHAR on them records nothing rather than failing the session.

It also needs the image to run the devtools helper. Verified support:

Browser	HAR
chrome, brave, MicrosoftEdge, opera, yandex	Yes
firefox, safari	No — no CDP endpoint

Each entry carries the request method, URL and headers, the response status, headers, MIME type and encoded size, and a total time derived from the CDP timestamps. Response bodies are not stored, which keeps a long session’s HAR to a size worth keeping.