July 8, 2026
Your AI bill has enemies
Show HN: Frugon – Find which LLM calls a cheaper model could handle (local, MIT)
A money-saving AI tool drops, and the comments instantly split between hype and eye-rolls
TLDR: Frugon is a free tool that checks your AI usage logs locally and shows where cheaper options could cut costs without sending your data away. Commenters were split between shrugging that the advice may be hard to act on and cheering the name like it was the thread’s real winner.
A new Show HN project called Frugon is pitching itself as the budget detective for AI apps: it scans your chatbot call logs on your own computer and tells you where you could switch to cheaper models and save cash. The creator’s big flex is privacy — your data stays on your machine, your keys go straight to your provider, and nothing gets sent back to them. In plain English: if your app is burning money talking to expensive AI, Frugon wants to show you exactly where the leak is.
But the real sparkle is in the reaction. One commenter basically looked at the whole thing and went, cute, but who can actually use this? The sharpest criticism came from cyanydeez, who argued this would be more interesting if it focused on using local AI setups and smarter multi-model workflows instead of just trimming the bill. Translation: for some power users, saving money isn’t juicy enough unless it comes with a grand master plan. That instantly turns the launch into a classic internet face-off: practical tool or niche spreadsheet energy?
And then, because no tech thread can survive on skepticism alone, another commenter cut through the seriousness with pure vibes: the name rules. Yes, Frugon got applause just for sounding right. So the thread ended up with the full tabloid cocktail: one camp asking whether this solves a real pain point, the other happily enjoying the branding while everyone side-eyes their rising AI bills.
Key Points
- •Frugon is a local, open-source tool that analyzes LLM call logs to estimate cost savings from switching or routing to cheaper models.
- •The tool can capture new logs with `frugon capture`, a local HTTP proxy that forwards provider calls unchanged and records them in JSONL format.
- •Frugon accepts logs in an OpenAI-style JSONL request/response schema, using provided token counts when available and its own tokenizer otherwise.
- •Its cost analysis runs fully locally without network access or LLM calls, while the optional `--measure` mode sends sampled prompts directly to the user’s own provider using their API keys.
- •Frugon can produce routing recommendations and export shareable HTML or Markdown reports summarizing estimated savings and quality trade-offs.