LiteLLM spend is per key. CostSlice is OpenAI cost per customer.
LiteLLM is a gateway. CostSlice is not. We are not a LiteLLM replacement. If you already run LiteLLM, keep it. Add CostSlice when the job is OpenAI cost versus plan price per customer.
LiteLLM is an AI gateway. It sits in front of many providers. It can issue virtual keys. It can apply budgets that stop calls. Teams use it to unify SDKs and to keep spend from running away. That is a real job. CostSlice is not that job.
CostSlice is a pass-through OpenAI proxy and a customer P&L. OpenAI only. You keep your OpenAI API key. We do not route. We do not cache. We do not issue a virtual key. We do not stop a tenant because a row is red. We count tagged usage and show cost versus plan price.
We are not a LiteLLM replacement. If you already run LiteLLM, keep it. If you need budgets that kill traffic, LiteLLM is built for that. If you need OpenAI cost per customer so finance can change a price, that is CostSlice. Some teams run both. That is fine. They are not the same box.
| LiteLLM | CostSlice | |
|---|---|---|
| Kind | AI gateway | Pass-through + customer P&L |
| Providers | Many | OpenAI only |
| Keys | Virtual keys common | You keep the OpenAI API key |
| Spend view | Per key / virtual key | Cost vs plan per customer |
| Budgets | Can stop calls | No tenant kill switch |
| Prompt bodies | Depends on setup | Never stored |
Fair: LiteLLM can show spend. The unit is usually the key. CostSlice’s unit is the customer you already invoice. Plan price sits next to OpenAI cost. That is a P&L, not a gateway dashboard.
Fair: LiteLLM budgets are a feature. CostSlice will not cap acme-health because the P&L is red. Starter will stop new proxy calls when the CostSlice plan hits 10,000 requests. That is our plan cap. It is not a per-tenant budget.
If the buying rule is “we will not replace the OpenAI API key with a virtual one,” CostSlice matches that rule. Authorization is forwarded. We never store it. See OpenAI cost without storing prompts.
Other gateways in this neighborhood: vs Portkey is the virtual-key catalog. Helicone logs and routes. The category page is not an AI gateway.
Point base_url at https://api.costslice.com/v1. Send x-cs-key and tag tenant, feature, env. Completions still go to OpenAI. We count. You read the row.
Missing tags do not drop the completion. Untagged still returns. We do not store prompt bodies. We do not train. Claude is not on CostSlice today. If LiteLLM is already routing Claude, that spend is not this P&L. Other models come later.
Starter includes 10,000 requests and seven days of the customer P&L. That is enough to see whether the gateway you already run still leaves finance without a row.
Starter $0 / Pro $79 / Scale $199, overage $6/100k. No credit card on Starter. Pricing on the homepage wins if a number here is stale.
Starter is free. No credit card. Operated by Operator Media Yard LLC, New Jersey. hello@costslice.com
Questions
- Is LiteLLM a gateway?
- Yes. LiteLLM sits in front of many providers. It can issue virtual keys and apply budgets that stop calls.
- Does CostSlice replace it?
- No. We are not a LiteLLM replacement. OpenAI only. No routing, cache, or virtual keys.
- Cost vs plan?
- CostSlice shows OpenAI cost next to the plan price you set for that customer. LiteLLM spend is typically per key or per virtual key.
- Tenant spend cap?
- We do not cap a tenant because a P&L row is red. Starter is a hard cap on the CostSlice plan. LiteLLM budgets can stop calls.
- Store prompts?
- CostSlice does not. We store usage metadata. Prompt and completion bodies are not written to disk.