AI Agent Request Inspector

Developers using Claude Code, Codex, or other AI coding agents can't see what prompts/requests are being sent to the model. This tool provides a local proxy + dashboard to inspect, log, and debug all API calls from coding agents in real-time.

Agentic coding tools shipped fast in 2024–2025, and observability for them is genuinely lagging — developers using Claude Code or Codex often have no visibility into what's actually being sent to the model, which matters both for debugging and for controlling costs. Fiddler and Charles Proxy exist as general HTTP inspectors, but neither surfaces LLM-specific context like token counts, prompt structure, or agent reasoning chains, so there's no clear incumbent in the AI-agent-specific layer. The $500–2k/mo revenue band is realistic for a solo dev tool with a small but sticky audience, though it implies the ceiling is low unless this becomes a team-level or enterprise compliance play. The biggest risk is that the AI coding agent vendors — Anthropic, OpenAI, Cursor — add native observability dashboards themselves, which would make a standalone proxy instantly redundant.