pporlock

I built pporlock to explore how developers and agents can inspect and transform browser traffic, and what safe, observable agent interaction with web infrastructure could look like. As a web developer, I’m often wondering what’s actually being sent and received by my web browser. This is a a combination of an https proxy, a set of modules that allow you to instrument and alter what’s flowing through, and an MCP interface that can inspect traffic and develop new modules and capabilities. A chrome extension makes it easy to manage when it’s in use and what it’s doing.

New modules are developed using a simple declarative language or with embedded python code. It comes with a set of examples and utilities:

• adblock - block traffic from known ad server domains
• cookie-banners - supress cookie warnings
• css-tamper - alter CSS elements as they go by
• json-tamper - alter JSON blocks
• ws-inspect- capture and inspect websocket traffic
• user-agent-switcher - emulate googlebot, an LLM scraper, or any other browser
• gpc-audit - send a global privacy control signal, and record the cookies that get set anyway

The MCP interface can monitor and capture traffic, develop a module to run the fix, and run the module against the shared traffic to make sure it worked.

The Agentic LLM integration was a great way to instrument the actual network traffic- but also the browser’s behavior. By selectively relaxing content security policies, I was able to pinpoint exactly what was breaking an external vendor’s site. I was able to diagnose a problem with our Global Privacy Controls and discover why we still had a cookie loading against user wishes. This last one had the potential to leave us with a significant GDPR violation, and the diagnosis was made much easier with this application.