Guide

Track AI spend across providers

4 min read

Most teams don't stand still on one provider — a support bot on one model, a coding agent on another, an embedding pipeline on a third. Each provider's own console only shows you its own spend, so the total picture lives nowhere until you build it yourself. Here's how to do that in a Meterlark project.

01

Create a project

Start a project for each product, team or workload you want to track separately — go to Projects, then New project. A project is the unit spend rolls up to; you can run multiple providers and multiple models inside the same one.

02

Log usage as it happens

Add a usage entry with the model, input tokens, output tokens and any cached tokens for a given day. Meterlark looks up the model in the live pricing dataset and computes the dollar cost automatically — you don't calculate it yourself, and it stays correct even if the provider changes its price later, because the entry stores the cost at the time it was recorded.

03

Import in bulk instead of entering rows one at a time

For any volume of usage, export a usage CSV from the provider's console and import it rather than typing entries by hand — see import OpenAI usage CSV or import Anthropic usage CSV for the exact steps.

04

Mix providers freely inside one project

Every entry is tagged with both the model and its provider, so a single project can hold OpenAI, Anthropic, Google and any other tracked provider's usage side by side. The dashboard totals across all of them without you separating spreadsheets per provider.

05

Read the unified total, not five separate consoles

Once entries exist across providers, the project dashboard shows total spend, a breakdown by model, and a spend trend over time — the view no single provider's own billing page gives you.

Once spend is tracked in one place, the next step is usually a budget with an alert — see set an AI budget that actually alerts.