Rate my idea and be brutally honest!
retained text
Would you pay for a "spending guard" for your AI API costs? Looking for honest feedback Hey everyone, I build AI products myself (two AI video apps) and keep running into the same fear: one bug or one agent stuck in a loop can burn a lot of money before anyone notices. The problem as I see it: - Agents can get stuck and repeat the same expensive call hundreds of times, often at night when nobody is watching - Provider spend limits are monthly. OpenAI and Anthropic don't offer a daily cap, so one bad night can eat your whole month's budget - At OpenAI, keys in the same project share one limit, so you can't cap a single agent or customer - A leaked API key basically works like a stolen credit card - Video and image APIs (fal, Replicate, etc.) cost cents to dollars per call, so a loop there gets expensive fast The idea: a simple guard that sits between your app and the AI providers: - Daily and hourly limits per key, per agent or per customer - Detects loops (the same request over and over) and stops them automatically - Instant alerts (Telegram, email, Slack) when spending spikes - One dashboard for OpenAI, Anthropic, Gemini, fal, Replicate and more - Setup: change one line in your code My questions: 1. Have you ever had a surprise AI bill? What happened, and how much? 2. How do you protect yourself today? 3. Would you trust a tool that sits in the middle of your API calls, or is that a dealbreaker? 4. Would you pay for this? Roughly how much per month?
document fetched 2026-11-12 00:29 · canonical https://reddit.com/r/SaaS/comments/1xtbicd/rate_my_idea_and_be_brutally_honest