Per-customer cost without running a proxy
LiteLLM is an open-source Python proxy you self-host — proxy plus PostgreSQL plus Redis — to track per-key spend. LLMeter gives you deeper per-customer cost and margin from a read-only API key. No proxy. No database. No infra to run.
No credit card required. Free plan never expires.
Why look past the proxy?
LiteLLM and LLMeter both track LLM spend. The difference is the infrastructure you carry to do it.
No infra to run
Spend tracking on LiteLLM means a proxy plus PostgreSQL plus Redis to deploy and keep healthy. LLMeter has nothing for you to run — connect a key.
Out of your request path
A Python proxy sits in front of every call, adding runtime overhead at scale and a failure mode. LLMeter reads usage out-of-band — zero latency.
Deeper attribution + margin
LiteLLM attributes per key, user, or team. LLMeter goes per-customer, feature, and environment — and adds cost-vs-revenue margin.
Set up in 3 steps
No SDK installation required. No code deployment. Just paste a key and go.
Sign up for LLMeter
Create a free account. No credit card required.
Paste your read-only API key
Go to your provider's dashboard, generate a read-only key, and paste it into LLMeter. That's it.
Keep LiteLLM for routing if you use it
Keep LiteLLM for routing/fallback if you use it — LLMeter does not replace your gateway, it adds per-customer cost and margin on top, with no extra infra.
LiteLLM vs LLMeter
A side-by-side comparison of what matters for per-customer LLM cost.
| Feature | LiteLLM | LLMeter |
|---|---|---|
| Setup method | Python proxy (self-host + route traffic) | Read-only API key + optional SDK tag |
| Sits in request path | Yes — proxy server | No |
| Added latency | Python runtime overhead at scale | Zero |
| Infra to run and maintain | Proxy + PostgreSQL + Redis | None |
| Attribution depth | Per-key / user / team (flatter) | Per-customer + feature + environment + margin |
| Cost vs revenue (margin) | No | Yes |
| Open source | Yes | Yes |
| Prompt/data access | Proxies all traffic | Never sees prompts or completions |
| Pricing | Open source + enterprise | Free / $19 / $49 |
No proxy. No database to babysit.
LiteLLM routes all your LLM traffic through a self-hosted proxy. LLMeter takes a fundamentally different approach.
LiteLLM (proxy-based)
- All LLM traffic routes through the proxy server
- Python runtime overhead on every call at scale
- Requires proxy + PostgreSQL + Redis to run
- The proxy sees all prompts and completions
LLMeter (API key-based)
- Reads usage data directly from provider APIs
- Zero impact on your LLM call latency or reliability
- No proxy, no database, no Redis — connect a key
- Never sees your prompts or completions
Migration FAQ
Do I have to give up LiteLLM to use LLMeter?
No. LiteLLM is a routing and fallback proxy; if you rely on those, keep it. LLMeter does not replace your gateway — it reads usage out-of-band through a read-only key and adds per-customer cost and margin on top. The two run side by side, and LLMeter adds no infrastructure of its own.
How does LLMeter calculate per-customer cost?
Per-customer cost is estimated from token counts. You tag each call with a customer_id using the open-source llmeter SDK or by POSTing to /api/ingest, and LLMeter applies current model pricing to those tokens. Provider-level totals — pulled through your read-only key — use the real billed amounts from the provider's usage API. So your top-line spend reconciles to the provider invoice, while the per-customer split is a token-based estimate.
Does LLMeter need a proxy, Postgres, and Redis like LiteLLM?
No. Running LiteLLM for spend tracking means standing up the proxy plus a PostgreSQL database and Redis, and keeping all of it healthy on your request path. LLMeter has no infrastructure for you to run — you connect a read-only key and it reads usage directly from the provider, out-of-band.
Is LLMeter free?
Yes — the Free plan supports 1 provider with 30-day retention, forever. The Pro plan ($19/mo) adds unlimited providers, alerts, anomaly detection, and 1-year retention. Team plans start at $49/mo.
Per-customer cost without the infra.
Keep LiteLLM for routing and layer LLMeter on top for cost and margin. Set up in 30 seconds — no proxy, no database, no code changes. Free forever for 1 provider.
No credit card required. Free plan never expires.