for people shipping AI · 7-day free trial

Know if your AI is making or losing money.

// theknowai = unit_economics(your_costs, your_price) → verdict

Every AI request has a real cost that flat SaaS pricing ignores. See your true gross margin, cost per action, and break-even runway — with a clear verdict in 30 seconds. No spreadsheets, no consultants.

Run free pricing check →

no charge · 7 days · cancel anytime

theknowai · margin verdict
#0042
gross margin
+68.4%
Healthy — room to grow
0 → 1,000 users
01anthropic.sonnet$0.0108
02openai.gpt4o-mini$0.0126
03vercel.hosting$20.00
04supabase.pro$25.00
05pinecone.starter$70.00
cost @ 1k users · 20 actions/mo
$9.38/user/mo
verified · 2026-06-21
2,000+
LLM models priced live
Daily
Price catalog refresh
< 30s
To your first verdict
4
Cost layers modeled

The problem

Flat pricing was built for software that costs nothing to run

AI isn't that software. Tokens are a real, variable cost of goods sold — and that's exactly where the money leaks.

Every action burns money

Classic SaaS costs almost nothing per request. Each AI action spends tokens — your cost scales with usage, not just signups.

Flat pricing bleeds margin

A power user can cost 5× a light user to serve. One flat price overcharges some and quietly loses money on the rest.

You find out from the invoice

By the time the model bill lands, the price is already live. The margin damage is done — and it's hard to walk back.

How it works

From raw costs to the right price, in one place

Three steps: connect your costs, get a defensible price, and ship — without rebuilding a spreadsheet every time a model price changes.

01

Map your costs

Add LLM calls, infra, APIs, and labor — or let AI scan your product and surface the costs you forgot.

02

See your real margin

We compute cost per action and cost per user from live model prices, then Claude Opus recommends the pricing model that fits.

03

Simulate & ship

Project 12 months, find your break-even month, and generate a Stripe-ready pricing page.

Free tool · no signup

The 30-second margin check

Pick your LLM, enter usage and price, see whether you're profitable per user — or quietly funding your customers.

01

Pick your model

Claude, GPT, Gemini — live token prices from 2,000+ models.

02

Enter usage & price

Tokens per request, requests per user, your monthly price.

03

See your real margin

Gross margin, break-even price, dominant cost source.

See it in action

What you actually get inside

Real product views — margin simulator, AI pricing engine, and agent run simulator. Drop in your numbers, see your real economics.

Margin simulator

12-month projections you can drag

See break-even month, gross margin, and MRR at the end of year one. Drag growth, churn, price, and actions/user sliders — chart updates live. Spot the month you stop losing money before you launch.

  • Cost-at-scale: $9.38/user at 1K users, full breakdown
  • Breakeven detection across optimistic / realistic / conservative
  • Persists your inputs — refresh keeps your scenario

MRR at M12

$48.2k

Gross margin

68.4%

Break-even

Month 6

12-month gross margin

in %
M1M3M6M9M12
Growth rate12%/mo
Churn5%/mo
Price/user$79
Actions/user150

Cost at 1K users

$9.38/user/mo

$5.60 fixed + $3.78 actions

AI pricing engine

Claude Opus picks your pricing model

Answer six questions about your product, market, and usage. Claude Opus recommends flat-rate, usage-based, credit-based, or hybrid — with specific tier prices and the reasoning behind each.

  • Specific dollar amounts per tier — not vague advice
  • Flags hidden traps (freemium with high LLM cost, etc.)
  • Generates Stripe-ready tier copy for your pricing page

Claude Opus 4.6 recommends

credit-based

High LLM cost variance + B2B SMB market → usage-based pricing aligns cost-to-serve with revenue. Suggested 3 tiers below.

Starter

$29

200 credits

  • Email support
  • 1 workspace

Pro

POP

$79

800 credits

  • Priority queue
  • Analytics
  • 3 workspaces

Business

$199

3,000 credits

  • SSO
  • Audit logs
  • Unlimited

Why credits, not flat rate?

Your heavy users cost 4.8× more to serve than your light users. A flat $79 plan would give 41% margin overall but lose money on heavy users. Credits align price with cost.

Agent run simulator

See what one agent run actually costs

Describe your agent as an ordered list of steps. Pick a model per step. See the per-step breakdown — and which step dominates. Catch the expensive LLM call before it eats your margin.

  • Per-step cost in mono — no hand math
  • Live 2,000+ model catalog from OpenRouter + LiteLLM
  • Margin verdict given the price you charge per run
1

Classify ticket

Claude Haiku

$0.0008
2

Retrieve KB context

GPT-4o-mini

$0.0012
3

Draft reply

Claude Sonnet 4.5

$0.0252
4

Quality check

GPT-4o-mini

$0.0004

Cost per run

$0.0276

4 steps · ~92% margin at $0.35

Breakdown

1. Classify ticket14%
2. Retrieve KB context22%
3. Draft reply56%
4. Quality check8%

Step 3 dominates at 56%.

Try Claude Haiku here to cut cost 70%.

Who it's for

Built for the people who own the price

If you decide what an AI product charges, you need to know what it costs to serve. That's the whole job.

Indie AI founders

Shipping a wrapper or AI SaaS solo. Price it right before launch — not after the first scary model bill.

Agent builders

Multi-step agents with a model per step. Find the one call that quietly eats most of your margin.

Product & growth teams

Setting or revising the price of an AI feature. Replace guesswork with defensible unit economics.

7-day free trial · Founding pricing

Try every feature free for 7 days

Start your free trial today — no charge for 7 days. After the trial, you're locked in at the founding rate, the lowest price we'll ever charge. Cancel anytime during the trial and pay nothing.

AI pricing engine

Claude Opus analyses your cost structure + market and recommends flat-rate, usage-based, credit-based, or hybrid pricing — with specific tiers and reasoning.

Cost discovery

Map every cost: LLM tokens, infra, third-party APIs, labor. We score what you might be forgetting and build your real cost-per-user automatically.

Agent run simulator

Describe your agent as an ordered list of steps. See cost-per-run, per-step breakdown, and margin given the price you charge.

Scenario simulator

12-month margin projection. Adjust growth, churn, conversion, price. Spot break-even month and total profit before you commit.

Pricing page generator

AI-written tier copy + headlines, ready for Stripe. Generate the pricing page you ship the same day.

Live LLM pricing catalog

2,000+ models across OpenAI, Anthropic, Google, DeepSeek, Meta, Mistral and more — refreshed daily from OpenRouter + LiteLLM.

Free vs founding

CapabilityFreeFounding
Free margin check (/check)UnlimitedUnlimited
ProjectsUnlimited
AI pricing recommendationIncluded
Cost structure engineIncluded
Agent run simulatorIncluded
Scenario simulatorIncluded
Pricing page generatorIncluded
Founding price locked in for life
Founding member

Skip the waitlist. Back the build.

Get instant access to the full pricing engine, simulator, and surveys. Founding price locked in for life.

$19/month

No charge for 7 days · Card required · Cancel anytime · Secured by Stripe

Secured by Stripe Cancel anytime Founding rate locked for life

FAQ

Questions, answered straight

Where do the cost numbers come from?
A live catalog of 2,000+ models from OpenRouter and LiteLLM, refreshed daily. You provide token volumes; we compute the dollars — input and output priced separately, reasoning uplift, and calls per action included.
How accurate are the estimates?
As accurate as your inputs. We don't hand you one optimistic average — we show P50 and P95 so you can see the spread, plus a cost-by-layer breakdown so you know exactly what's driving it.
My agent isn't a simple chatbot. Does it still work?
Yes. Model any agent as an ordered list of steps with a different model per step. The simulator shows per-step cost and flags the step that dominates your margin.
Why not just use a spreadsheet?
A spreadsheet goes stale the day a model price changes. theknowai pulls live prices, does the token math, and turns it into a pricing recommendation in 30 seconds — not an afternoon.
Is my cost and pricing data private?
Yes. Your projects are scoped to your account with row-level security. We don't sell your data and we don't train on it.
Do I need a credit card to try it?
The free margin check needs no signup at all. The 7-day full-product trial uses Stripe Checkout and requires a card — but you are not charged until day 7, and you can cancel anytime before then and pay nothing.
What happens after the 7-day trial?
You're billed at the founding rate — the lowest price we'll ever charge — locked in for life. Cancel during the trial and you pay $0.
Which model providers are covered?
OpenAI, Anthropic, Google, DeepSeek, Meta, Mistral and more — 2,000+ models in total, kept current daily.

// stop pricing on vibes

Find out what your AI really costs

Run the free 30-second check now, or start the 7-day trial and get the full toolkit — AI pricing engine, agent simulator, and scenario modeling.

Run free pricing check

no charge · 7 days · cancel anytime