Use cases
Costr surfaces the hidden cost problems in your AI stack. Then it gives you the data to act on them.
The scenarios below are illustrative, not real customer data. Costr is new. These are the patterns I built it to catch.
PART ONE / WHAT COSTR SURFACES
The hidden cost problems Costr makes visible.
Multi-tenant agents firing LLM calls per customer session.
The problem
You run AI agents on behalf of customers. Each customer's agent fires multiple LLM calls per session. Total bill keeps growing but you can't tell which customer is responsible.
What Costr does
Per-customer cost across every agent invocation. Margin per customer in real time.
Heavy users burn 10x light users, everyone pays the same.
The problem
Your product creates per-session AI output — notes, transcripts, drafts. Some users burn 10x what others do. Pricing is flat. Margin is silently negative on a fraction of accounts.
What Costr does
Cost per session, cost per active user, which power users are pulling margin into the red.
On-demand generation. Some users generate 100 articles, some generate 5.
The problem
Customers generate content on demand. The distribution is wildly uneven. Every account looks the same in your billing dashboard.
What Costr does
Per-account cost breakdown. Which accounts to reprice, which to keep, which to escalate.
Your OpenAI bill keeps surprising you. You don't have time to instrument cost tracking.
The problem
You're building solo. Paying out of pocket. Your AI bill keeps creeping up. You'd love to know which feature or customer is costing what, but you don't have a free weekend to build it.
What Costr does
Plug-and-play visibility. Free tier covers $25/mo of tracked spend — enough to run real numbers without a credit card.
OpenAI for one feature, Anthropic for another, Google for a third. Bills never reconcile.
The problem
You're routing different workloads to different providers. The bills come from three sources and never reconcile against your customer list.
What Costr does
Unified per-customer view across all three providers in one dashboard.
PART TWO / WHAT YOU DO WITH THE DATA
Costr is the diagnostic. Here's what teams actually do with it.
Reprice the 4% of customers eating 38% of your margin.
What Costr gives you
A ranked list of every customer by AI cost vs MRR, updated in real time.
What you do with it
Move unprofitable customers to usage-based tiers or higher seats. Heavy users typically accept repricing — they expect it.
Ship a new AI feature without flying blind on cost.
What Costr gives you
Per-feature cost attribution. Real cost per call, broken down by customer.
What you do with it
Price the feature based on what it actually costs to deliver. Set caps. Know your break-even point before you announce.
Stop guessing what your AI bill will be next quarter.
What Costr gives you
Per-customer, per-feature, per-provider spend trends over 30/60/90 days. Run-rate projections grounded in real usage.
What you do with it
Build an AI cost forecast like any other COGS line. Brief your board on the cost trajectory at current growth rate.
Know which signups are actually going to be profitable.
What Costr gives you
Cost-per-customer in the first 7 days. Early signals on heavy vs light usage patterns.
What you do with it
Reprice, cap, or let unprofitable customers churn before they drain margin for six months.
Route the right workload to the cheapest provider per customer.
What Costr gives you
Per-customer cost broken out by provider (OpenAI vs Anthropic vs Google). See which provider is most expensive for which workload.
What you do with it
Route low-stakes customers to cheaper models. Reserve premium models for enterprise tier. Move workloads between providers based on cost data, not guesses.
Free under $25/mo tracked spend. One-line integration.