FAQ

Is this just another OpenRouter?

No. OpenRouter resells tokens with a markup and runs on its own credits. OpenLLM runs on the accounts you already own, resells nothing, and adds no markup. The headline difference is that OpenLLM can drive your tools from a consumer subscription, not only a metered API key.

Do I pay OpenLLM for tokens?

No. Every request bills to your own provider account at your provider's price. The OpenLLM plan covers the platform features (vault, fallback, dashboard, plugins), not the model usage. See Pricing.

Can it really use my Claude or ChatGPT subscription?

Yes, through the local daemon. The subscription runs on your own machine against the vendor's official CLI, using your own login. Those credentials never route through the OpenLLM cloud. See Providers and Compliance.

What happens when a provider is down?

Your fallback chain catches it. OpenLLM moves to the next provider in the tier and returns the first good answer, without dropping your request.

What are lite, plus, and ultra?

Tier aliases you use as the model name. Each resolves to a real provider and model through your fallback chain, so you can reorder providers without changing any code. See Concepts.

How safe are my keys?

Provider API keys are stored only as ciphertext and decrypted in memory per request, so a database dump reveals nothing usable. Subscription credentials stay on your machine. Full detail on Security.

What is the recovery phrase, and what if I lose it?

It is the master secret behind your encrypted vault, shown once at setup. Everything you connect is encrypted under it, and OpenLLM keeps no copy, so save it in a password manager immediately. If you lose it there is no reset link: re-keying the vault permanently deletes the credentials and saved memories encrypted under the old phrase. Treat it like a crypto wallet's seed phrase.

Which tools can I connect?

Claude Code, Codex, the Kimi CLI, and Raycast have one-line installers today, plus the openllmc CLI and the local daemon. Anything else that speaks OpenAI or Anthropic works by pointing its base URL and key at your gateway. See Connect your tools.

Does it run on Windows?

The web dashboard works anywhere, but the CLI and daemon are macOS and Linux only right now.

Can I use OpenLLM on more than one computer?

Yes. Give each machine its own sk-llm-… key and, if you use subscription providers, its own local daemon. A plan covers up to three devices on one subscription. See Use multiple devices.

Can my team share one account?

Not yet. OpenLLM is single-user today. One account owns its keys, data, and usage. The five-device allowance is for your own machines (laptop, desktop, server), not seats for other people.

Can I make my agents cheaper and faster?

Yes. The OpenLLM plugin adds semantic code search and persistent memory so agents read fewer files and stop re-deriving what you already told them, so fewer tokens go into each request on any model. In-flight compression of large documents (PDF, DOCX) is coming to shrink requests further.