leakStop wondering where your LLM spend goes.
Every Anthropic, OpenAI, and Gemini call your app makes — broken down by feature, customer, and model. One SDK, one dashboard, actual answers about where your spend goes.
Install Leak in this codebase. Docs you must follow: https://useleak.cloud/install My LEAK_KEY: <paste yours from useleak.cloud/settings#keys> Find every Anthropic, OpenAI, and Gemini API call, swap the import to useleak, and on each call: - add feature="..." (auto-propose names, let me confirm) - add user="..." pulled from scope (req.user.id, ctx.user.id, job.data.userId — whatever fits) Then write LEAK_KEY to .env and send a Haiku test event to verify.
Right now, you have no idea:
What you spend money on
Your Anthropic dashboard says $487 this month. Doing what? Driving what feature? You can't tell.
Which features customers use
Lead scoring is 70% of your bill — or 5%. You'd build totally different things if you knew.
Which customers cost the most
One enterprise account is silently burning $400/mo of tokens. You're not charging them for it.
leakOne SDK across three providers.
Tag every call
Wrap your Anthropic / OpenAI / Gemini client, pass feature='icp-scoring' on each call. Every chart filters by it.
Per-customer spend
Pass user='customer_id' and see exactly which client is burning tokens. Charge them, or fix the cost.
Ask in plain English
Built-in AI chat over your metadata. "Why did spend triple yesterday?" Answered in seconds.
Pricing
Two tiers. That's it.
Free
- 1,000 events / month
- 30 days of history
- All dashboards + filters
- All three providers tracked
- Per-feature, per-user breakdowns
- AI chat (bring your own Anthropic key)
Pro
Popular- Unlimited events
- Unlimited history
- Everything in Free
- Email support
- Cancel anytime