Use your own AI provider key. Your key never touches our servers.
On Plus, Pro and Business you can point the SnoutData assistant at your own OpenAI, OpenRouter, or Anthropic key. The assistant's chat and edits go straight from your machine to that provider, and the key and those prompts never pass through SnoutData's servers. AI completion in the editor is the exception: it needs a fill-in-the-middle model, which your provider may not offer, so if you switch it on (it is off by default) it uses SnoutData's completion model through our gateway.
What BYOK is, and who it's for
By default the assistant runs on SnoutData's managed gateway: we serve the models, meter usage, and bill it to your plan. BYOK swaps that out for your own provider account. You paste an API key, pick a model, and the assistant talks to that provider directly. You pay the provider for what you use, at their rates, on your own account.
It is for people who already have a provider account, want a specific model, or have a policy that says prompts must go only to an approved vendor (in that case, leave AI completion off: it needs a fill-in-the-middle model, so it runs on SnoutData's). BYOK is part of Plus, Pro and Business.
Where your key lives
Your key is stored in your operating system's keychain: Windows Credential Manager, the macOS Keychain, or libsecret on Linux. It is encrypted at rest by the OS and is never written to a config file, a log, or our servers. It stays on the machine where you entered it.
Where requests go
On the BYOK path the assistant's chat and edits bypass the SnoutData gateway. The app builds the prompt locally, the same way it does on the managed path (schema linking, grounding, and all), then sends it straight to your provider's endpoint using your key. We are not in the middle: we never receive the key, the prompt, or the response. Because of that, BYOK turns off the parts that only the gateway provides, such as the shared wallet and usage metering. One feature does not move to your key: AI completion, the ghost text in the editor. It needs a fill-in-the-middle model, and we cannot count on your provider offering one, so it always uses SnoutData's completion model through our gateway. It is off by default; when you switch it on, the text around your cursor, the relevant schema and some of your earlier queries on that connection are sent to us.
Why it's a paid feature
A fair question, so here is the honest answer. The app, the prompt pipeline, and the schema grounding are identical on both paths. BYOK does not give you a better assistant. What it unlocks is the direct-to-provider routing and the plumbing around it: keychain storage per provider, model discovery, and the bad-key handling that makes it pleasant to use. That work, and keeping it maintained across three providers, is what sits behind the Plus, Pro and Business line.
Per-provider setup
Add a key under Settings, AI, in SnoutData Studio, then pick a model. Where to get a key:
- OpenAI: create a key at platform.openai.com/api-keys. The app lists your account's chat models.
- OpenRouter: create a key at openrouter.ai/keys. One key reaches many models across vendors.
- Anthropic: create a key at platform.claude.com. Pick from the Claude models, or type the exact model id you want.
Each provider has an optional base URL field too, so you can point at a proxy, Azure OpenAI, or a self-hosted gateway that speaks the same API.
Get started
BYOK is part of Plus, Pro and Business. Upgrade your plan, then download or update the app and add your key under Settings, AI.