Burnstack vs Helicone
Helicone is an open-source AI gateway that logs every LLM request with its estimated cost. Burnstack reads what your vendors actually billed, across AI and everything else, without touching your code.
The short version
- Choose Helicone if you need per-request logs, cost per user and gateway routing; you want rate limits based on cost in your request path.
- Choose Burnstack if you want the billed truth, including usage that never went through a proxy; AI is one of several bills you need to watch; you don't want to change your code.
At a glance
| Burnstack | Helicone | |
|---|---|---|
| What it is | Desktop app (Mac; Windows soon) that reads your vendors' billing APIs | Open-source AI gateway and LLM observability |
| How it gets data | Billing APIs | Requests routed or logged through it |
| Cost figures | What you were billed | Estimates from model and token counts |
| Code changes | None | Route traffic through it or add logging |
| Cost per request or user | — | ✓ |
| Price | Free forever; Enterprise for teams | Free 10k requests/month; Pro $79/month plus usage |
Who owns the data
Helicone runs in its cloud or self-hosted (Apache-2.0). Your requests flow through its gateway, which either manages provider keys for you or uses your own (docs). Costs are best-effort estimates from model detection and token counts, not billing data (docs).
With Burnstack, API keys stay in your computer's keychain and your spend history is a file on your disk. You can export or delete all of it any time, and none of it is sent to us. What owning your data means.
Which bills each one reads
| Vendor | Burnstack | Helicone |
|---|---|---|
| Vercel | ✓ | AI SDK tracing, not billing |
| Neon | ✓ | — |
| Supabase | ✓ | — |
| OpenAI | ✓ | Request logging |
| Anthropic | ✓ | Request logging |
| GitHub billing | ✓ | — |
| OpenRouter | ✓ | Request logging (source) |
| DigitalOcean | ✓ | — |
| Twilio | ✓ | — |
| Stripe fees | ✓ | — |
| Cloudflare | ✓ | Not found |
| Netlify | ✓ | — |
| AWS | ✓ | Bedrock request logging, not billing (source) |
| Datadog | ✓ | — |
| Sentry | ✓ | — |
| Cursor | ✓ | — |
| PlanetScale | ✓ | — |
| Azure | ✓ | Azure OpenAI request logging, not billing (source) |
| Google Cloud | ✓ | Vertex AI request logging, not billing (source) |
| MongoDB Atlas | ✓ | — |
| Railway | ✓ | — |
Burnstack also reads Render, Heroku, Akamai (Linode), Vultr, Fastly, Runpod, xAI, Together AI, fal.ai, Runway and Upstash (new in 0.9.0). We haven't checked Helicone's coverage of those yet, so they aren't in the table.
Pricing
Hobby is free (10k requests, 7-day retention), Pro $79/month plus usage, Team $799/month, and Enterprise includes on-premise deployment (Helicone pricing).
Burnstack is free forever. Enterprise (several accounts per vendor, monthly reports, team features) is priced per team. Pricing.
Where Helicone is better
- Per-request cost, latency and error logs.
- Cost by user or segment with custom properties.
- Gateway routing, fallbacks and cost-based rate limits.
- Alerts on cost, errors and latency to email or Slack.
Where Burnstack is better
- Reads actual invoices, so it also catches usage that bypassed the proxy: a leaked key, a script, a teammate's test.
- Hosting, database and GitHub costs in the same view.
- No code change and nothing in your request path.
A note on Helicone's status
Helicone was acquired by Mintlify in March 2026 and moved to maintenance mode (Mintlify). In October 2026 its site said it's "no longer part of Mintlify" with details to come (helicone.ai). Check its current status before adopting it.