Skip to content
CLEVYA CLEVYA Docs

BYOAK - Bring Your Own AI Key

Your key, your provider, your bill

CLEVYA does not resell tokens. You plug in your own key with the provider of your choice - Anthropic, OpenAI, Mistral, or a local model via Ollama - and you pay your consumption directly to that provider. CLEVYA bills the use of the platform, not the AI.

This is what BYOAK means: Bring Your Own AI Key.

Why it matters

  • No intermediary on your tokens. The cost of the AI is transparent: it is your provider’s bill, which you see directly.
  • Model independence. CLEVYA is designed to be model-agnostic. The coupling point between the agent loop and the model is deliberately minimal, which makes it possible to switch providers without rewriting the logic.
  • Sovereign fallback. You can order a fallback chain that ends on a local model (Ollama). If the cloud providers are unavailable, the agent continues on site - the anonymized data does not go out.

Where you plug in your key

You enter your key in the dashboard, not in a config file or an environment variable. Once your instance is set up (first-run screen), go to Settings → LLM Providers: add a key (Anthropic, OpenAI, Mistral, or the address of a local Ollama), test it, and set it as default. Each agent uses the workspace default key, or its own key if you assign one.

The key is encrypted on your server (AES-256-GCM, one derivation per row) before being stored; it never appears in clear text, neither in logs nor in an API response. The same place serves the agent mode and the proxy mode: a single entry, no double configuration.

Tracking consumption

CLEVYA measures the tokens consumed per agent (input, output, and cache read/write, billed at distinct rates). There is no provider interface that returns a reliable “remaining balance”: CLEVYA therefore shows your real local consumption, never an invented balance. A budget per agent and per day can be set to bound the spend.

Going further