betool lets you choose, model by model, between two ways of accessing an LLM:
- Managed models — the platform operator makes shared LLM accounts available. You use them without connecting a key; consumption is debited in credits, according to a per-token rate for input and output. No model is free.
- BYOK (Bring Your Own Key) — you supply your own API keys (or the URL of your private model). Inference costs are billed directly by your provider; betool takes nothing on tokens.
In regulated environments, BYOK is the recommended option: your agreement (DPA) with the provider remains the sole contractual chain, your billing is readable line by line, and you revoke the key without going through us. See BYOK — why it is non-negotiable in regulated industries.
Supported providers
| Provider | Models | Mode |
|---|---|---|
| OpenAI | Current GPT generation | Public API |
| Anthropic | Claude (Opus, Sonnet, Haiku) | Public API |
| Gemini | Public API | |
| Ollama | Open-weight models, custom models | Self-hosted, or Ollama Cloud |
| LM Studio | Local models | Self-hosted |
| OpenAI-compatible endpoint | vLLM and other servers exposing the OpenAI API | Self-hosted (custom URL) |
The managed model catalog follows the current generation of Claude, GPT and Gemini, as well as open-weight models served by Ollama Cloud (Kimi, GLM…). The exact, up-to-date list is visible in the admin's model selector.
Step 1 — Retrieve your API key
Follow your provider's procedure:
- OpenAI → platform.openai.com/api-keys
- Anthropic → console.anthropic.com/settings/keys
- Google → aistudio.google.com
- Ollama / LM Studio / OpenAI-compatible endpoint → the URL of your instance (e.g.
http://ollama.internal:11434)
Step 2 — Register the key in betool
- In the admin panel, open Settings → Providers.
- Click + Add an account.
- Select the provider, paste the API key (and the URL for a self-hosted model).
- Give it a readable name (e.g. "OpenAI - prod").
The key is never returned by the API once saved. If you need to regenerate it, you will have to update it manually. GET routes only return has_api_key: true | false.
Step 3 — Choose your models
Each agent node in your pipelines then references the account and the model to use — the same account can serve several models.
For example, one OpenAI account can serve a high-end model for primary agents and a lighter model for fast classification.
Shared accounts
Shared LLM accounts (the is_shared flag) are the ones that carry the managed models. They are visible to organisations without connecting a key, and resolution always favours your own accounts (own > shared).
Pricing
See Pricing for metering details. Reminder: in BYOK mode, we do not charge for LLM tokens — only for our orchestration. On a managed model, tokens are debited in credits.