What LiteLLM is
LiteLLM is an open source proxy that speaks the OpenAI API and forwards requests to OpenAI, Anthropic, xAI, and many other providers. Teams run it to give every tool one address, hand out virtual keys instead of real provider keys, set budgets, and log usage. It is free, and it is good software.
FUSR.ai runs on LiteLLM
To be plain about it: FUSR.ai is a hosted LiteLLM proxy. Your FUSR.ai key is a LiteLLM virtual key. Your provider API keys are stored encrypted in that proxy, not in the app database, and are never shown again.
On top of LiteLLM, FUSR.ai adds the parts you would otherwise build yourself: routing with fusr/auto, prompt caching that stays warm, daily and monthly caps per key, a savings report in plain words, and a one command setup for your tools.
Running it yourself, or not
Both are fair choices. Here is the honest trade.
| Run LiteLLM yourself | FUSR.ai | |
|---|---|---|
| Cost on top of tokens | Nothing, plus your server and database | Nothing on tokens. Free plan, or a flat Pro or Team plan |
| Setup | A server, a Postgres database, config, and upgrades | Paste your API keys, then npx fusrai setup |
| Routing | You write the rules and the hooks | fusr/auto, with rules you can read and change |
| Caps and alerts | LiteLLM budgets, set up by you | Daily and monthly caps per key, alerts at the amount you choose |
| Reports | Raw logs and spend data | A savings report in plain words, plus a weekly digest on Pro and Team |
| Control | Every LiteLLM feature and provider | Chat and text completions for Claude, OpenAI, and xAI |
When to run it yourself instead
- You need providers or endpoints FUSR.ai does not route, like embeddings, images, or Anthropic's own message format.
- Requests must stay inside your own network.
- You already run infrastructure and a proxy is one more small service.
Otherwise, a hosted proxy saves you the server, the database, the upgrades, and the on call. For how FUSR.ai compares with OpenRouter, Requesty, and Portkey, see the AI router guide.
Moving from your own LiteLLM
Tools that talk to your LiteLLM today already speak the OpenAI API, so moving is a base URL and a key. Connect your provider keys on FUSR.ai, run npx fusrai setup, and ask for fusr/auto, or name a model with fusr/claude, fusr/openai, or fusr/grok. Keep your own proxy running until you are happy, then turn it off.