TokenOps — Per-Customer LLM Cost & Gross Margin for AI

Unit economics platform for AI companies

See per-customer profit Ask your agent why.

One line of code wraps Anthropic, OpenAI, Bedrock, Google, Vercel AI, Azure, and more. TokenOps captures every event, reconciles your monthly vendor invoice, and breaks cost down by the customer who incurred it — so you actually know which accounts are profitable.

30MCP tools

4Workflow tools

6Vendor wrappers

<5 minnpm install → first event

Live · last 5 customers

May 2026 · running

Revenue − Cost = Margin · matched to invoices, customers & models

Customer LLM cost Revenue Margin
Acme Robotics $4,820.12 $9,000.00 +46.4%
Brightline $1,291.00 $1,900.00 +32.1%
Northwind Labs $2,104.55 $2,400.00 +12.3%
Verdant Inc. $580.40 $1,500.00 +61.3%
Halcyon $3,210.00 $2,000.00 −60.5%
import { TokenOps, wrapAnthropic } from "@tokenops/sdk";
// One line. Every call now attributed.
const client = wrapAnthropic(newAnthropic(), tokenops);
await client.messages.create({ customerId });
// ← attributes the spend

The wedge

Per-customer cost is the question your spreadsheet can't answer.

AI-native companies pay a handful of vendors, get one invoice each, and have no shared schema for cost. That makes the most important question — is this customer profitable? — a research project every month.

Invoice mystery

The Anthropic invoice doesn't say which customer cost you what.

Vendor bills aggregate at the workspace level. By the time it lands, you've lost the join between spend and the customers who incurred it.

Anthropic · April 2026 INV-44-2089

Event In Out
claude-sonnet-4-5 $1,236.00 $1,320.00
claude-haiku-4-5 $960.00 -

Total: $3,516.00

Rate drift

Effective rate ≠ list rate. You find out at month end.

Effective $ / 1M tokens April 2026
List: $3.00
Effective: $2.41
Δ −19.7%

Your CEO wants per-customer gross margin. By Friday.

One spreadsheet, six tabs of CSVs, two SQL joins, and a promise to ‘firm it up next month.’ This is most AI startups today.

What it does

One SDK. One reconciliation engine. One MCP your agents already speak.

TokenOps slots in at the LLM client and stays out of your way. You wrap once, attribute by customer, and the rest — capture, normalize, reconcile, expose — happens for you.

01 Wrap any LLM client

One-line wrappers for Anthropic, OpenAI, Bedrock, Google, Vercel AI, Azure, and more. We never proxy your traffic — calls go directly to the vendor.

02 Reconcile invoice → events

Every event is priced against captured tokens. Monthly vendor invoices are diffed line by line so you see effective rate, drift, and savings per (vendor, period, customer).

03 Answer via MCP

A 30-tool MCP catalog plus 4 composite workflows. Connect Claude Desktop, Cursor, or Codex and ask finance questions in plain English — the agent reaches for the right tools.

Workflow tools

Four composite tools for the four finance questions you actually ask.

Each workflow stitches a dozen MCP reads into one structured answer — the kind of multi-step reasoning agents do well, served as a single tool so it's deterministic.

How it works

Install. Wrap. See. Around five minutes.

01 Install the SDK

Drop @tokenops/sdk (or the Python package) into the service that holds your LLM clients. Edge runtimes welcome.

$ yarn add @tokenops/sdk
$ export TOKENOPS_TENANT=…

02 Wrap your LLM client

One line per vendor. Pass customerId on each call — that's the join key for per-customer cost.

const client = wrapAnthropic(newAnthropic(), tokenops); // every call now attributed

03 See per-customer cost

Open the dashboard, hit /api/cost, or ask your agent. Cost lands within seconds; the monthly vendor invoice is reconciled when it arrives.

$48,210.55

LLM cost across 142 customers

Pricing

Start free. Talk to us when you scale.

Try TokenOps on your own telemetry for 30 days, no card required. When you outgrow the trial, we'll size a plan to your volume.

Common questions, candid answers.

How is this different from Helicone or other tools? TokenOps is a finance layer for unit economics. We share the capture step, but the rest of the product is reconciliation, gross margin, and the MCP layer that lets an agent answer ‘is Acme profitable?’

Do you proxy our LLM calls? No. The SDK wraps your client in your runtime. Calls go directly to the vendor.

What does the SDK actually capture? Per request: model, input/output tokens, vendor-reported usage, latency, and your customerId.

How does invoice reconciliation work? Each month you forward the vendor invoice to TokenOps, and we match it against captured events.