Providers & Models

Bring your own key

Point Flowly at your own provider account instead of Flowly Cloud. Flowly bills nothing in BYOK mode — the provider charges you directly for what you use. Add a key in one command and switch to it instantly.

Why BYOK

You already pay a provider and want to use that quota.
You want a specific model Flowly Cloud doesn't carry.
You want billing to sit entirely with the provider, not Flowly.
You run your own model server (vLLM, an OpenAI-compatible proxy).

BYOK is a CLI / terminal feature

Bring-your-own-key providers are configured from the CLI and the terminal chat (/integrations, /provider). The desktop app's save flow writes the Flowly Cloud slot — it won't touch your BYOK keys, so the two coexist safely.

Supported providers

Provider	Slug	Models	Get a key
OpenRouter	`openrouter`	Hundreds — Claude, GPT, Gemini, Llama, Kimi, …	openrouter.ai/keys
Anthropic	`anthropic`	Claude (Opus, Sonnet, Haiku)	console.anthropic.com
OpenAI	`openai`	GPT family	platform.openai.com
xAI	`xai`	Grok (via API key)	console.x.ai
Google Gemini	`gemini`	Gemini family	aistudio.google.com
Groq	`groq`	Fast open models — also powers voice transcription	console.groq.com
Zhipu	`zhipu`	GLM family	z.ai
vLLM / custom	`vllm`	Any OpenAI-compatible endpoint you host	self-hosted

New to BYOK? Start with OpenRouter

One OpenRouter key reaches almost every model — Claude, GPT, Gemini, Llama, Kimi, and more — so you can try different models without signing up for each provider separately.

Adding a key

One-shot from the CLI — save a key and pin it as the active provider in a single command:

bash

flowly setup byok openrouter            # prompts for the key (hidden input)
flowly setup byok anthropic --key sk-ant-...   # or pass it inline
flowly setup byok groq --no-set-active  # save it but don't switch to it yet

Valid slugs: openrouter, anthropic, openai, xai, gemini, groq. The key is saved to ~/.flowly/config.json and applies on the next message.

From the terminal chat — for a guided experience with a live connection probe, open the integrations manager:

bash

/integrations    # pick a provider card, paste the key, Flowly probes it live
/provider        # then switch between every provider you've configured

Once the key is in, choose the model with /model. See Switching providers & models for the full switcher workflow.

Grok without an API key

If you have a SuperGrok or X Premium+ subscription, you can use Grok in Flowly by signing in with that subscription instead of pasting an xAI API key — no per-token billing. It's set up like an integration rather than a BYOK key:

xAI Grok — sign in with your subscription →

Self-hosted & custom endpoints

Running your own model server, or pointing at an OpenAI-compatible proxy? Use the vllm slot and set its apiBase to your endpoint. Any provider's apiBase can be overridden the same way — handy for gateways and regional endpoints. The full schema is in the configuration reference.

json

{
  "providers": {
    "active": "vllm",
    "vllm": {
      "apiKey": "EMPTY",
      "apiBase": "http://localhost:8000/v1"
    }
  },
  "agents": { "defaults": { "model": "your-served-model-id" } }
}