Helicone dashboard
Helicone logs every LLM request automatically when you route through the AI Gateway or Helicone proxy — cost, model, tokens, and latency appear in the Requests tab within seconds.
Hub: Tokenminning with Helicone.
This page answers helicone dashboard, helicone cost tracking, and helicone analytics.
First request
- Sign up at helicone.ai
- Generate an API key at API Keys
- Send a request through
https://ai-gateway.helicone.ai - Open Requests — your call appears with full metadata
What to review weekly
| Column | Action |
|---|---|
| Model | Which slug dominates spend? |
| Cost | Total $ by feature or user |
| Prompt tokens | Input bloat signal |
| Completion tokens | Verbose output signal |
| Latency | Slow models cost more wall-clock time |
Filter by date range and export for finance reconciliation.
Proxy mode (existing apps)
If you already call OpenAI directly, add Helicone as a proxy without changing your SDK:
const client = new OpenAI({
apiKey: process.env.OPENAI_API_KEY,
baseURL: "https://oai.hconeai.com/v1",
defaultHeaders: {
"Helicone-Auth": `Bearer ${process.env.HELICONE_API_KEY}`,
},
});You get logging without switching to the AI Gateway.
Cost attribution
Pass custom properties in headers:
defaultHeaders: {
"Helicone-Property-Feature": "code-review",
"Helicone-Property-Env": "prod",
}Group Requests by property in the dashboard.
Spending guardrails
- Review Requests tab weekly — top 3 models by cost
- Export before month-end for reconciliation
- Compare Helicone cost to provider invoice — should match at 0% markup
- Trace one expensive request before changing model defaults
Related
Last updated on