Skip to main content
The AdsCrawl Codex plugin adds browser-rendered page reading and screenshots to OpenAI Codex. It bundles an adscrawl-browser skill and a Python 3.9+ command that uses only the standard library. It calls the AdsCrawl API using your ADSCRAWL_API_KEY and does not bundle credentials.

​
Features

  • Render a URL as Markdown, HTML, or structured article JSON.
  • Capture a full-page or viewport PNG screenshot.
  • Select a managed country route and navigation wait strategy.
  • Use AdsCrawl CDP sessions for interactive browser work through your project’s browser automation library.

​
Requirements

​
Install

Install the plugin from the public GitHub repository:
Start a new Codex task after installation so the bundled skill is loaded. To add a local checkout while developing:

adscrawl-codex on GitHub

View source and releases.

​
Authenticate

After installing the plugin, ask Codex to connect your AdsCrawl account, or run the auth command directly:
The command asks for your account email and a project name, then prints a link for you to approve. It never asks for your password or prints the resulting API key. Alternatively, set ADSCRAWL_API_KEY in your environment. The plugin will reuse an existing environment variable. The auth flow saves your key privately at ~/.config/adscrawl/codex.env with 0600 permissions.

​
Direct commands

The skill invokes plugins/adscrawl/scripts/adscrawl.py. You can also run it directly:

​
Render a page

--format accepts markdown, html, or json (structured article).

​
Capture a screenshot

Run python3 plugins/adscrawl/scripts/adscrawl.py <command> --help for the available options. API failures are not retried automatically.

​
CDP sessions

For interactive browser work (clicking, filling forms, multi-step flows), use AdsCrawl CDP sessions through your project’s browser automation library. See Remote CDP browsers and Playwright automation.
Plugin requests can consume AdsCrawl credits. See Credits and limits.