Get an API key

Clyde works with a hosted provider when you bring a key, and with a model on your own machine when you don't. Pick one, get the key from its page below, and give it to Clyde in a minute.

Prefer a page to a terminal? Run clyde luv bonnie, open Providers, and Bonnie walks you through the same three steps and tests the key before saving it.

Add a key in three steps

Get the key. Open your provider's key page from the table below, sign in, create a key and copy it. Treat it like a password.
Run clyde login. Pick the provider from the list (type to filter), paste the key when it asks (nothing shows as you type), then pick a model. Clyde saves the key and your choice.
Start working. Run clyde in your project. Switch models any time with /model.

Prefer the shell? Export the provider's variable instead, for example export OPENAI_API_KEY=your-key (the names are in the table). A key exported in your shell always wins over a saved one. Inside Clyde, /login does the same as clyde login.

Providers

Thirteen hosted providers have a built-in place for their key. Pricing and free allowances belong to each provider, so check their pages; Clyde does not sell keys or credits.

Provider	Get the key	Environment variable
Anthropic	platform.claude.com/settings/keys	`ANTHROPIC_API_KEY`
OpenAI	platform.openai.com/api-keys	`OPENAI_API_KEY`
Google (Gemini)	aistudio.google.com/apikey	`GEMINI_API_KEY`
OpenRouter	openrouter.ai/keys	`OPENROUTER_API_KEY`
Mistral	console.mistral.ai/api-keys	`MISTRAL_API_KEY`
NVIDIA	build.nvidia.com/settings/api-keys	`NVIDIA_API_KEY`
DeepSeek	platform.deepseek.com/api_keys	`DEEPSEEK_API_KEY`
Cerebras	cloud.cerebras.ai/platform	`CEREBRAS_API_KEY`
GLM (Zhipu)	open.bigmodel.cn/usercenter/apikeys	`GLM_API_KEY`
MiniMax	platform.minimax.io/console/access	`MINIMAX_API_KEY`
Cloudflare Workers AI	dash.cloudflare.com/profile/api-tokens also needs your account ID (`wrangler whoami` shows it); `clyde login` asks for it	`CLOUDFLARE_API_TOKEN`
Pollinations	enter.pollinations.ai	`POLLINATIONS_API_KEY`
Ollama Cloud	ollama.com/settings/keys only for Ollama's hosted models; a local Ollama needs no key	`OLLAMA_API_KEY`

No key: models on your machine

Ollama: Install it from ollama.com/download and run ollama serve. Clyde finds it on its own.
oMLX: MLX models on Apple Silicon. Clyde installs it from its source and starts it the first time you pick one, after asking.

Inside Clyde, /models local searches ollama.com and Hugging Face for models that fit your machine, rates each one, and downloads the one you pick.

Any other service

Together, Fireworks, Groq, vLLM, LiteLLM, a proxy of your own: anything that speaks the OpenAI or Anthropic API works. Run clyde login and choose custom. Clyde asks for a name (for example together), the protocol, and the base URL, which is the part before /chat/completions (OpenAI) or /v1/messages (Anthropic), then the key. A server that needs no key takes an empty one.

Good to know

Where it is kept: Saved keys live in ~/.clyde/keys.json, readable only by you. /doctor inside Clyde checks that.
Check it worked: clyde --list-models lists every model you can use right now, from each connected provider.
Change or remove it: Run clyde login again to replace a key, or clyde logout openai to forget one.
Keep it private: Never paste a key into a prompt, a chat or a commit. Clyde redacts saved keys from its trace files.

Provider pages change. If a link no longer lands on the key page, sign in to the provider's dashboard and look for "API keys".