nable - One usage meter for every AI provider you pay for

hey product hunt, im chandan, solo-founder of nable! As I was building nable, I kept running out of 5hr limits, weekly limits, etc. And a lot of the tasks ended up not being super critical, but running out would prevent me from working further, so I built nable guard! nable guard puts a budget on your AI coding agent, and can stop it when it crosses one. uvx nable guard install Your agent then checks its own budget before it starts something expensive. You set the budget once and pick what happens when you cross it: notify you, or stop the agent. Default is notify. Nothing blocks unless you asked it to. Underneath the guard is the meter that powers it. One command, real numbers, no key and no account: uvx nable ai-budget Claude Code is read from session logs already on your machine. OpenAI, Anthropic, OpenRouter, LiteLLM, Modal, Together, Replicate, Cohere and Mistral connect with a key. A dollar figure is meaningless on a flat subscription. You don't care what your tokens "cost," you care how close you are to the limit that's about to cut you off mid-task. So nable measures flat plans in usage and metered plans in dollars. Two honest limits. Token counts are exact; dollar figures are list-price estimates, not your invoice. And the hard stop only truly blocks inside Claude Code, the one client with a pre-tool hook. Everywhere else it runs over MCP, so the agent has to cooperate. All local. Apache-2.0, nothing uploaded!