What is OpenLLM

OpenLLM is a bring-your-own-credentials LLM gateway. You connect the AI accounts you already have (Claude, ChatGPT, Grok, and others), you get back one key, and every tool you use points at that single key.

Behind that one key, OpenLLM routes each request to a real provider using your own credentials. It never resells tokens and never adds a markup. You pay your providers directly, exactly as you do today.

The problem it solves

If you live in AI coding agents like Claude Code, Codex, or the Kimi CLI, your setup today usually looks like this:

• You pay a flat monthly subscription (Claude Max, ChatGPT Plus, SuperGrok).
• On top of that, you may also be paying metered API rates to drive those agents, paying for the same model twice.
• Your keys are scattered across env vars, config files, and half-working shell scripts, one per tool.
• When a provider rate-limits you or hits capacity, the task just stops. There is no automatic fallback.

OpenLLM collapses all of that into one endpoint, one key, and one dashboard.

What you get

• One endpoint, your accounts. A single API backed by the plans and keys you already own. Any tool works with it, whichever provider it was built for.
• Automatic fallback. Define a tier like ultra or plus and an order of providers. If one is down, rate-limited, or over context, OpenLLM steps to the next one without dropping your request.
• Encrypted credentials. Provider keys are stored only as ciphertext. See Security.
• No markup. OpenLLM never resells tokens. You pay your providers directly.
• One dashboard. Every request across every provider, with cost and failure rates in one place — including what your subscription usage would have cost at metered API rates, so you can see exactly what you save.

Who it is for

The core user is a developer who runs AI coding agents every day and already pays for at least one premium AI subscription or API plan. If that is you, the Quickstart gets you to a working request in about five minutes.