Launching today
Guardrail by NeatProxy
Full visibility into your AI spend, before it goes rogue
4 followers
Full visibility into your AI spend, before it goes rogue
4 followers
We burned real money running AI agents and didn't know why until it was too late. That's the whole reason Guardrail exists. It's a local proxy for Claude Code and Codex that shows you where every dollar goes, system prompt, tools, that MCP server you forgot about, before you spend it. When you're about to blow a budget, Guardrail blocks the request. Not after. Before. Built for developers who'd rather see the bill coming than explain it later.
Hey Product Hunt! 👋
I'm Jitte, co-founder of NEAT. We built Guardrail because we got burned by our own AI spend while running a previous company: runaway agent loops, surprise bills, no way to see it coming until the invoice hit.
Guardrail is a local proxy that sits between your dev tools (Claude Code, Codex) and the AI providers. It blocks requests before they fire once you hit a budget threshold, instead of just reporting on the damage after the fact. A few things we're proud of:
Pre-send blocking, not post-hoc reporting
Local-first: your data stays on your machine, cloud sync is opt-in
Hidden Costs view that shows exactly what you're paying for (including MCP servers you forgot you connected)
A live status line so you see today's spend under every Claude Code prompt
We've been running this in closed beta with 35 users across several teams, and today we're opening it up more broadly.
Would love your feedback, especially if you're already burning real money on Claude Code, Codex, or other agentic tools. What's the scariest AI spend surprise you've had?