For B2B SaaS with AI in the product
This customer is unprofitable.
Change one base_url. Tag tenant, feature, and env. Costslice shows which accounts and features eat LLM margin — in time to cap them before month-end.
No card for Hobby. We never store prompts or completions.
Tenant P&L · this month
Red rows: LLM cost above the customer’s plan price
| Tenant | LLM cost | Plan price | Margin | Requests |
|---|---|---|---|---|
| acme-healthlosing money | $4,812 | $2,400 | -101% | 18,240 |
| harbor-support | $978 | $900 | -9% | 6,402 |
| lumen-chat | $612 | $3,200 | 81% | 44,110 |
| northwind-analytics | $441 | $1,800 | 76% | 3,901 |
| pebble-docs | $188 | $600 | 69% | 2,104 |
| untagged | $96 | — | set price | 412 |
01
Point the SDK at us
Keep your OpenAI key. Set base_url to api.costslice.com/v1 and send x-cs-key so we know which workspace the traffic belongs to.
02
Tag tenant / feature / env
Required headers, 128 characters, opaque ids — not emails. Forget a tag and we still complete the call, then show untagged so you can fix it.
03
Read the P&L
Spend this month vs last, by model. Tenants and features sorted by dollars. Optional plan price → margin %. CSV when you need a board pack.
Pricing that fits a product org, not a platform team.
Hobby is free without a card. Overage on Team and Scale is $6 / 100k requests.