Burnstack vs LiteLLM
LiteLLM is an open-source LLM gateway you run yourself, with virtual keys and budgets that actually block requests. Burnstack doesn't block anything; it shows the whole bill across vendors and warns you early.
The short version
- Choose LiteLLM if you need to enforce LLM budgets per key, user or team; you route many models through one gateway.
- Choose Burnstack if you want to see what you were billed across AI, hosting, database and CI; you don't want to run and secure a proxy server.
At a glance
| Burnstack | LiteLLM | |
|---|---|---|
| What it is | Desktop app (Mac; Windows soon) that reads your vendors' billing APIs | Self-hosted LLM proxy with budgets |
| Can block spending | — | Yes: budgets reject requests |
| How it gets data | Billing APIs | Traffic through the proxy, priced from its cost map |
| What you run | A desktop app | A server and a database |
| Price | Free forever; Enterprise for teams | Open source free; Enterprise by quote |
Who owns the data
You host LiteLLM yourself, so provider keys live in your proxy's configuration or secret manager and spend is written to your own database (docs). That's real data ownership too, at the cost of running a server. Spend is priced from LiteLLM's model cost map rather than reconciled with invoices.
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 | LiteLLM |
|---|---|---|
| Vercel | ✓ | As an AI routing provider, not billing |
| Neon | ✓ | — |
| Supabase | ✓ | — |
| OpenAI | ✓ | Proxied provider |
| Anthropic | ✓ | Proxied provider |
| GitHub billing | ✓ | — |
| OpenRouter | ✓ | Proxied provider (source) |
| DigitalOcean | ✓ | — |
| Twilio | ✓ | — |
| Stripe fees | ✓ | — |
| Cloudflare | ✓ | Not found |
| Netlify | ✓ | — |
| AWS | ✓ | Bedrock as a proxied provider, not billing (source) |
| Datadog | ✓ | — |
| Sentry | ✓ | — |
| Cursor | ✓ | — |
| PlanetScale | ✓ | — |
| Azure | ✓ | Azure OpenAI as a proxied provider, not billing (source) |
| Google Cloud | ✓ | Vertex AI as a proxied provider, 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 LiteLLM's coverage of those yet, so they aren't in the table.
Pricing
The open-source proxy is free, including virtual keys, spend tracking and budgets. Enterprise (SSO, audit logs, support) is quoted annually (LiteLLM pricing).
Burnstack is free forever. Enterprise (several accounts per vendor, monthly reports, team features) is priced per team. Pricing.
Where LiteLLM is better
- Hard budgets per key, user, team and model that reject requests.
- Budget alerts to Slack, Teams or webhooks.
- Routing and fallbacks across 100+ providers.
Where Burnstack is better
- Shows billed cost, including usage that never went through the proxy.
- Hosting, database and GitHub costs in the same view.
- Nothing to deploy or keep secure.
In March 2026, two LiteLLM PyPI releases were briefly compromised in a supply-chain attack; the official Docker images weren't affected (LiteLLM). Pin versions whatever you run.