Per-customer cost without a gateway
Bifrost by Maxim AI is an AI gateway — every LLM call routes through it to enforce virtual keys and budgets. LLMeter gives you the same per-customer view, plus cost-vs-revenue margin, from a read-only API key. No gateway, no latency hop, no infra to run.
No credit card required. Free plan never expires.
Why look past the gateway?
Bifrost and LLMeter both attribute spend per customer. The difference is what sits in your request path.
No gateway to run
Bifrost is self-hosted infrastructure on the hot path. LLMeter reads usage out-of-band — there is nothing to deploy, scale, or keep up.
Out of your request path
A gateway is one more hop on every LLM call and one more thing that can fail. LLMeter never touches traffic, so it cannot add latency or break a call.
Cost plus margin
Bifrost gives per-customer budgets. LLMeter adds the revenue side — per-customer unit economics, so you see margin, not just spend.
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.
Run LLMeter alongside Bifrost
Keep Bifrost in place if you want gateway enforcement — LLMeter reads usage out-of-band, so it runs alongside without touching your request path.
Bifrost vs LLMeter
A side-by-side comparison of what matters for per-customer LLM cost.
| Feature | Bifrost | LLMeter |
|---|---|---|
| Setup method | AI gateway (route traffic through Bifrost) | Read-only API key + optional SDK tag (no gateway) |
| Sits in request path | Yes — all LLM calls route through it | No — reads usage out-of-band |
| Added latency | Gateway hop (must run the gateway) | Zero (never in your request path) |
| Per-customer attribution | Yes (virtual keys, 4-tier budgets) | Yes (customer_id via SDK / ingest) |
| Cost vs revenue (margin) | No | Yes — per-customer unit economics |
| Infra to run and maintain | Yes — self-host the gateway | None — connect a key |
| Open source | Yes (Apache 2.0) | Yes |
| Prompt/data access | Proxies all traffic | Never sees prompts or completions |
| Pricing | Open source + enterprise (contact sales) | Free / $19 / $49 |
No gateway. No request-path risk.
Bifrost routes all your LLM traffic through its gateway. LLMeter takes a fundamentally different approach.
Bifrost (gateway-based)
- All LLM traffic routes through the gateway
- A gateway hop adds latency to every call
- You self-host and maintain the infrastructure
- The gateway proxies 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 infra to run — connect a read-only key
- Never sees your prompts or completions
Migration FAQ
Does LLMeter sit in my request path like Bifrost does?
No. Bifrost is an AI gateway — every LLM call routes through it, which is also how it enforces virtual keys and budgets. LLMeter never proxies traffic. It uses a read-only API key to read usage and billing data out-of-band, so it adds zero latency and can never break an LLM call. You can keep Bifrost running and add LLMeter alongside it.
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.
What does LLMeter add that Bifrost's budgets don't?
Bifrost gives you per-customer budgets and enforcement at the gateway. LLMeter adds the revenue side: per-customer unit economics, so you can see cost against revenue (margin), not just spend. It does this without a gateway, without code in your request path, and without infrastructure to run.
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 margin without the proxy hop.
Run LLMeter alongside Bifrost for a week and compare. Set up in 30 seconds — no gateway, no infra, no code changes. Free forever for 1 provider.
No credit card required. Free plan never expires.