Is your AI feature profitable?

One bill tells you what AI cost you. Hereling tells you who spent it, on which feature, and whether they paid you more than they cost.

See the live demo

We are letting people in from the waitlist. The demo opens without an account.

No provider keys stored. No prompts stored. No access to your database.

With Hereling

Revenue+$2,480.00
AI cost−$1,903.40
Revenue − AI cost+$576.60

Without Hereling

Your invoice$1,903.40
OpenAIClaudeGeminiDeepSeekGrokKimiMiniMaxMistralOpenRouter

One line to connect

Point your AI calls at Hereling and add two headers: who the user is and which feature they are using. Everything else about your request stays exactly as it was.

  • We never store your provider key

    It travels in your own request, we forward it, and it is gone. Not in our database, not in our logs.

  • We never store prompts or answers

    Only numbers: model, tokens, cost, duration, and the two labels you send.

  • We never touch your database

    We don't ask for a connection string. Everything comes from the proxy and your payment webhooks.

your app · one change
baseURL: "https://api.openai.com/v1"
baseURL: "https://api.hereling.com/v1/openai"

headers: {
  "x-hereling-key":     "hl_live_…",
  "x-hereling-user":    userId,
  "x-hereling-feature": "chat"
}

Works the same for OpenRouter, Anthropic and DeepSeek: swap the last part of the URL.

How it works

  1. 1

    Route your AI calls through Hereling

    One base URL, two headers: your user id and the feature name. Your provider key stays in your own request.

  2. 2

    Point your payment webhooks at us

    Stripe, Paddle or RevenueCat. We give you a URL, you paste it. No API keys, no database access.

  3. 3

    See who is making or losing you money

    Recorded revenue and AI cost per user and plan, plus AI cost per feature. Taxes, fees and other expenses are not deducted.

The log at the top of this page is everything Hereling keeps: who called, which model, how many tokens, what it cost. Never the prompt, never the answer.

See who's costing you money

Every user you have, with what they paid you next to what they cost you. Most of them are fine. The few that aren't are the reason your margin moves.

This month412 users
Revenue+$2,480.00
AI cost−$1,903.40
Revenue − AI cost$576.60

3 users cost more than they pay

  • user_8821Free−$41.20
  • user_1043Pro−$4.90
  • user_5520Pro−$2.15

Then do something about it

Seeing the loss is half of it. The other half runs in the proxy you already route through, so nothing changes in your app.

Cap a user

A monthly limit per user. Choose availability or strict protection for supported calls.

$5.00 / month · user_8821

Switch the model

Move a heavy user to a cheaper model. You pick it here, not in your code.

gpt-6-astra gpt-5.6-luna

Get told early

A budget blown, a feature that suddenly costs more, revenue with no user attached.

  • user_8821 is over budget
  • “summarize” costs 5× what it usually does
  • $58 of revenue has no user attached

Start free. Pay when you act.

Seeing the numbers costs nothing. Capping a user or moving them to a cheaper model is the paid part.

Free

$0

forever

  • 10,000 requests / month
  • 7 days of history
  • Users, features and plans views
  • Budget alerts
Join the waitlist
Most popular

Pro

$25

per month, per project
plus tax in your country

  • Unlimited requests
  • 30 days and this month views
  • Weekly email digest
  • Cap a user's monthly spend
  • Switch a user to a cheaper model
  • Actions from Claude Code via MCP
Join the waitlist

We are letting people in from the waitlist. Every account starts on Free. Upgrade a project to Pro from its Settings page when you want the actions. Pro is billed monthly through Paddle, which acts as the merchant of record. Tax is added at checkout. Cancel anytime.

Find out before your next invoice does.

10,000 requests a month on the free plan once you are in. No card needed.