Docs

LLMeter is an OpenAI-compatible proxy. Change your base URL, use your LLMeter API key, and keep your code exactly as-is.

1. Get your key

Sign up, then copy your key from the dashboard. It looks like llm_live_....

2. Add a provider key

In Settings, paste your OpenAI, Anthropic, or OpenRouter key. It's encrypted at rest (AES-256-GCM). LLMeter forwards calls using your key — you keep your own billing relationship with the provider.

3. Point your SDK at LLMeter

from openai import OpenAI

client = OpenAI(
    base_url="https://meter.aiskillhub.info/v1",
    api_key="llm_live_xxxxxxxxxxxxxxxxxxxx",
)

resp = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello"}],
    extra_headers={
        "x-llmeter-feature": "onboarding-email",
        "x-llmeter-user": "user_123",
    },
)

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://meter.aiskillhub.info/v1",
  apiKey: "llm_live_xxxxxxxxxxxxxxxxxxxx",
});

await client.chat.completions.create({
  model: "gpt-4o",
  messages: [{ role: "user", content: "Hello" }],
}, {
  headers: {
    "x-llmeter-feature": "onboarding-email",
    "x-llmeter-user": "user_123",
  },
});

curl https://meter.aiskillhub.info/v1/chat/completions \
  -H "Authorization: Bearer llm_live_xxxxxxxxxxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -H "x-llmeter-feature: search" \
  -H "x-llmeter-user: acct_42" \
  -d '{"model":"claude-3-5-haiku-20241022","messages":[{"role":"user","content":"hi"}]}'

Attribution headers

• x-llmeter-feature — which feature made the call (e.g. support-summary).
• x-llmeter-user — which end-user/account, so you can see cost per customer.

Auto-routing

Turn on routing in Settings and pick a quality floor. LLMeter sends each prompt to the cheapest model that clears the bar across the providers you've configured. Routed calls return an x-llmeter-routed-to header and are tagged in the dashboard, with estimated ₹ savings.

Supported models

All OpenAI and OpenRouter chat models (OpenAI-compatible passthrough), plus Anthropic Claude via automatic translation. Pricing for known models is used for cost + routing; unknown models pass through and are tracked by token count.