LLMeter
Now live · Free 100k tokens/mo

See where your LLM token spend goes — and cut it.

A drop-in proxy for OpenAI, Anthropic & OpenRouter. Logs every call, shows cost per feature and per user, and auto-routes prompts to the cheapest model that meets your quality bar. Built for Indian dev teams.

Change one line of code. Get your first cost breakdown in 5 minutes.

One line. That's the whole integration.

Point your existing OpenAI SDK at LLMeter and keep using it exactly as before. We meter the call, attribute the cost, optionally route it cheaper, and forward it to your provider with your own key.

  • 1. Add your provider key once (BYOK — encrypted at rest).
  • 2. Swap your base_url and use your LLMeter key.
  • 3. Tag calls with feature/user headers — watch the dashboard fill up.
app.py
from openai import OpenAI

client = OpenAI(
    base_url="https://meter.aiskillhub.info/v1",  # ← only change
    api_key="llm_live_xxxxxxxxxxxxxxxxxxxx",       # your LLMeter key
)

client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Summarise this ticket"}],
    extra_headers={
        "x-llmeter-feature": "support-summary",
        "x-llmeter-user": "acct_4821",
    },
)

Everything you need to control LLM cost

No enterprise FinOps suite. Just the three things that actually move your bill.

Cost per feature & per user

Tag any call with two headers. Instantly see which feature and which customer is burning your token budget — not just one big monthly number.

Auto-route to the cheapest model

Set a quality bar; LLMeter routes each prompt to the cheapest model that clears it — GPT-4o → Llama-3.1 or DeepSeek when it's good enough. Typical teams cut 30-60%.

Budget alerts in ₹

Set a monthly ₹ budget per project. Get warned at 80% — before the bill, not after. No surprise invoices in USD.

Your keys stay yours

BYOK. Provider keys are AES-256-GCM encrypted at rest. We never train on, sell, or log your prompt bodies by default.

Drop-in, OpenAI-compatible

Change one line — your base URL. Works with the OpenAI SDK in Python, Node, anything. No rewrite, no new client.

₹-first, not FinOps bloat

Built for Indian SaaS founders and small teams. Razorpay + UPI billing, rupee dashboards, pricing that fits an indie budget.

Why now

“OpenRouter's raise put model-cost optimization on every founder's radar. But the tools are built for enterprise FinOps teams with USD budgets. Indian teams need rupee-aware routing that just works.”

— Why we built LLMeter

Built to sit in your hot path — safely

The questions every senior dev asks before routing traffic through a proxy.

Do you store my prompts?

No. We persist only metadata — model, token counts, cost, and your feature/user tags. Prompt and completion bodies are never written to disk.

What about latency?

A few milliseconds on top of your provider's own latency. Streaming responses are passed through untouched, token by token.

Who holds my keys?

You do. Provider keys are BYOK, AES-256-GCM encrypted at rest, and you keep your own billing relationship with OpenAI/Anthropic.

What if LLMeter is down?

Your provider keys are yours — revert your base_url and you're calling the provider directly again in one line. No lock-in.

Simple, ₹-first pricing

Start free. Upgrade when tracking pays for itself — which is usually in week one.

Free

For trying it out and side projects.

₹0/mo

$0 — no card required

  • 100,000 tokens tracked / month
  • Cost per feature & per user
  • All providers: OpenAI, Anthropic, OpenRouter
  • 7-day call history
  • 1 project
Most popular

Pro

For teams shipping LLM features in production.

₹1,499/mo

≈ $18 — billed monthly via Razorpay (UPI / cards)

  • Unlimited tokens tracked
  • Auto-routing to the cheapest model
  • Budget alerts in ₹
  • Unlimited history & projects
  • CSV export & priority support

You bring your own provider keys and pay OpenAI / Anthropic / OpenRouter directly. LLMeter only bills for the tracking + routing layer — never for your tokens.

Stop guessing where your AI bill goes.

Five-minute setup. Free forever for small projects.