Know which customers actually make you money.

Per-customer AI cost, true margin, and hard budgets for teams shipping LLM features.

Start free Read the docs

app.margint.dev/customers

Live demo

Total MRR

$8.4K/mo

AI cost

$4.2K

+18.4%

Gross margin

49.8%

−3.2pp

Customers underwater

2/ 14

Attention · biggest margin drain · 30d

HyperBrief pays you $599 a month and costs you $842 in inference. Every call they make is a small donation from your runway.

−41%

Margin

View customer →

CustomerSourceMRRAI costMarginState

HBHyperBriefops@hyperbrief.aiStripe$599$842−40.6% Bleeding

CCContextCloudbilling@contextcloud.devStripe$449$4128.2% Thin

CDClerkDeskfinance@clerkdesk.ioStripe$299$21428.4% Thin

NLNorthwind Labsops@northwind.ioStripe$1299$25780.2% Healthy

FAFenway AIhello@fenway.aiCSV$999$18681.4% Healthy

TLTessellateteam@tessellate.coStripe$849$13284.5% Healthy

Your AI bill is one number. Your margin is hundreds.

OpenAI sends one invoice. Stripe sends MRR. In between — the customers bleeding you, and the ones quietly funding your roadmap.

Langfuse and Braintrust know which prompt ran. Not who paid for it.
Stripe knows what they pay you. Not what you pay OpenAI.
Tokens don't fit a spreadsheet. The cost sheet is wrong before you save it.

CustomerSourceMRRAI costMarginState

HBHyperBriefops@hyperbrief.aiStripe$599$842−40.6% Bleeding

CCContextCloudbilling@contextcloud.devStripe$449$4128.2% Thin

CDClerkDeskfinance@clerkdesk.ioStripe$299$21428.4% Thin

NLNorthwind Labsops@northwind.ioStripe$1299$25780.2% Healthy

FAFenway AIhello@fenway.aiCSV$999$18681.4% Healthy

TLTessellateteam@tessellate.coStripe$849$13284.5% Healthy

Install in three minutes. See margin in real time.

TypeScript SDK, Python SDK, or one curl. Cost, margin, and budgets appear in the dashboard. No proxy, no latency hit.

TypeScript and Python wrappers, or hit the ingestion endpoint from any language.
Tokens, model, cost. Never prompts or completions.
OpenAI, Anthropic, Google, Mistral, Cohere, Bedrock, local models.

import { Margint } from '@margint-ai/sdk'
import OpenAI from 'openai'

const m = new Margint({ apiKey: process.env.MARGINT_KEY! })

// Wrap any provider client — tracking is automatic
const openai = m.wrap(new OpenAI(), {
  customerId: user.id,
  feature: 'chat'
})

await openai.chat.completions.create({
  model: 'gpt-4o',
  messages: [{ role: 'user', content: prompt }]
})

Your dashboard is for humans. Your API is for agents.

Every screen has a matching endpoint. Scripts, CI, and any MCP-aware agent read margin in one call.

Stable REST. Bearer auth. Rate-limited. OpenAPI shipped.
Drop one JSON into Claude Desktop or Cursor. Ask: "who's losing me money?" No dashboard required.
guardedCall() lets an agent check its budget before it overspends yours.

$ curl -H "Authorization: Bearer $MARGINT_KEY" \
    https://app.margint.dev/api/v1/customers | jq '.data[0]'

{
  "externalId": "cust_hyperbrief",
  "mrrCents": 24900,
  "costCents": 84200,
  "marginPercent": -238.2,
  "marginState": "negative"
}

The financial layer for AI products.

One product, one job. Answer "which customers make me money" before your investors ask.

Per-customer margin table

Every customer, sorted by margin. Red at the top means they're eating your lunch.

Cost by feature

Prove your agent costs 14× more than chat. Break down by feature, provider, model.

Budget guard

Wrap calls in guardedCall() — block customers over limit before the provider bills you.

Margin alerts

Email when a customer's margin drops below your threshold or workspace average tumbles.

Stripe revenue sync

Connect once. We pull MRR per customer and do the subtraction you've been doing by hand.

LiteLLM-backed pricing

400+ models, always fresh. Mirrored from LiteLLM's canonical price database.

Stop guessing. Ship with margin.

Free up to 100k events / month. No credit card. Three minutes to install.

Start free Read the docs