AI providers

Open Settings → AI Providers to configure the providers listed below. Their models then appear in the model selector in the chat composer.

Built-in providers

Provider	Type	Notes
Kora	Managed	Flash through `https://api.kora-ai.download/v1`; requires Kora Premium and a valid Kora key.
ChatGPT Plus/Pro	OAuth	Use your existing OpenAI subscription directly. No API key needed.
ByteDance ModelArk	BYOK (`bytedance`)	Two modes: standard (Seed models) and coding-plan (Seed + GLM, DeepSeek, Kimi).
Kimi Code	BYOK (`kimi-code`)	Kimi coding models through `https://api.kimi.com/coding/v1`. Temperature on Kimi models is normalized to a minimum of `1.0`.
OpenCode	CLI	Dynamic discovery via `opencode models --verbose`. Kora inherits your existing OpenCode config (API key, models, preferences) automatically.
Devin	CLI	Connects through your installed Devin CLI. Kora inherits your Devin login, streams its reasoning, and lists its slash commands in the palette.

Custom providers

Add an OpenAI-compatible provider from the models.dev catalog, with a manual definition, or by connecting a local inference server. See Custom providers.

Method	Best for
Catalog	Providers in models.dev; Kora fills the base URL, models, and capabilities
Manual	Providers not in the catalog, or full control over configuration
Local	Ollama, omlx, LM Studio — discovery via `GET /v1/models`

Per-model capabilities

Each model entry exposes four capability toggles: Tools, Streaming, Vision, Reasoning. Edit them at any time from the provider detail view. Per-model temperature is also configurable — see Per-model temperature.

Troubleshooting

A built-in provider is missing → close and reopen Settings; built-in entries cannot be deleted.
A configured provider has no models → refresh its catalog or verify the provider credential.
Kora shows an invalid key → reconnect the key using Connect Kora Premium.