DevSpeak supports Bring Your Own Key (BYOK) — a feature that lets you connect your own LLM API credentials to use custom models for translation.
| Provider | Tier Requirement | Configuration |
| ----------------- | ---------------- | -------------------------------------------- |
| OpenAI | All tiers | API key + model identifier |
| Anthropic | All tiers | API key + model identifier |
| Gemini | All tiers | API key |
| OpenAI-Compatible | All tiers | API key + custom endpoint + model identifier |
Gemini is a BYOK-only provider: it is available when you connect your own key, and is not part of DevSpeak's own model selection.
OpenAI-Compatible covers OpenRouter (which has its own preset) and the other providers DevSpeak can reach: DeepSeek, Mistral, Cerebras, NVIDIA, and Alibaba Cloud Model Studio. DevSpeak calls your endpoint from its own servers, so only endpoints on its allowlist can be saved. A model server on your own machine, such as Ollama or LocalAI, works only in a self-hosted DevSpeak deployment.
1. Navigate to Settings → Keys & Providers.
2. Select your provider card.
3. Enter your API key (stored encrypted server-side) and, except for Gemini, the exact model identifier.
4. Click Test Connection to verify the credentials.
5. Click Save Provider. Translations then route through your chosen provider unless you explicitly pick a DevSpeak model from the model selector. Leaving the selector on Auto also uses your provider: Auto omits the model pin, and an unresolved pin falls through to your stored credential before the router picks a managed model.
BYOK is available on every subscription tier, including FREE. It is always optional — it changes which model executes your translation, not whether you can translate at all.
| Status | Meaning |
| -------------- | --------------------------------------- |
| active | Credentials validated and in use |
| invalid | Last validation failed — check your key |
| unconfigured | No custom provider set up |
| validating | Connection test in progress |