AI Agent Plugin

RoxyProxy ships with an AI agent plugin. Once installed, your AI agent knows all RoxyProxy commands, filters, API endpoints, and common debugging workflows -- so you can ask it to inspect traffic in natural language.

Installation

The plugin is currently available for Claude Code. Add it from the Claude Code plugin marketplace:

/plugin marketplace add rvanbaalen/roxyproxy

Then open the plugin browser and install it:

/plugin

Support for other AI coding agents (Cursor, Copilot, Windsurf) is planned. In the meantime, any agent can use RoxyProxy directly via the CLI and REST API.

What It Provides

After installation, your AI agent can help you:

  • Start and stop the proxy and configure HTTPS interception
  • Write roxyproxy requests queries with the right filter flags
  • Use the REST API to query captured traffic programmatically
  • Debug failing API calls by inspecting captured request/response pairs
  • Set up system-wide proxy routing on macOS

Just ask your AI agent anything about intercepting or inspecting HTTP traffic and it will use its knowledge of RoxyProxy to help.

Example Workflows

Debug a failing webhook

Tell your AI agent something like: "The Stripe webhook is failing. Debug it."

The agent will start the proxy if needed, query RoxyProxy for recent requests matching the webhook endpoint, inspect the request headers and response body, and pinpoint what went wrong -- all without you writing a single filter command. See the full Debug a Failing Webhook recipe for the step-by-step walkthrough.

Find why your API returns 422

Ask: "Why is my API returning 422?"

The agent filters captured traffic by status code 422, inspects the request body that triggered the error, reads the response error message, and explains what validation failed and how to fix it. The Debug 422 Errors recipe covers this workflow in detail.

Capture iOS traffic

Ask: "Help me set up the proxy to capture traffic from my iPhone."

The agent walks you through starting the proxy, finding your network address, configuring the iOS proxy settings, installing the CA certificate on the device, and enabling full trust -- step by step. Follow the Capture iOS Traffic recipe for the complete guide.

Agent Output Mode

For optimal AI consumption, the CLI supports a --format agent output mode. This produces structured output that any LLM can parse efficiently when querying traffic programmatically — it works with Claude Code, Cursor, Copilot, or any other AI agent. Use it when you want your agent to process large amounts of captured traffic:

roxyproxy requests --format agent --host api.example.com
← HTTPS Interception Architecture →