Skip to main content
Cloud providers give you access to the most powerful frontier models available today — Claude, GPT-4o, Gemini 2.5 Pro, and many more. You bring your own API key for each provider you want to use; Luminy stores it locally on your device and uses it directly when making inference requests. No key ever passes through Luminy’s servers.

​
Adding a cloud provider API key

1

Get your API key

Visit the provider’s developer console and create an API key (see the per-provider instructions below).
2

Open Settings in Luminy

Click the gear icon or press ⌘, / Ctrl+,.
3

Select the provider section

Scroll to the provider you want to configure — for example, Anthropic or OpenAI.
4

Paste your key and save

Paste the API key into the field and click Save. The provider’s models appear in the model selector immediately.
API keys are stored locally on your device and are never transmitted to Luminy’s servers.

​
Provider details

Anthropic

Anthropic’s Claude models are best-in-class for code understanding, long-context reasoning, and instruction following.Get your key: console.anthropic.comRecommended models:
Model IDNotes
anthropic:claude-opus-4-5Most capable, best for complex tasks
anthropic:claude-sonnet-4-5Balanced speed and capability
anthropic:claude-haiku-3-5Fastest and most cost-efficient
Prompt caching: Luminy takes full advantage of Anthropic’s prompt caching, which significantly reduces costs on long conversations and large codebase sessions. Repeated context — system prompts, file contents, project rules — is served from cache rather than counted as new input tokens each turn.
OpenAI’s GPT and o-series models offer strong all-around coding performance and the broadest ecosystem support.Get your key: platform.openai.comRecommended models:
Model IDNotes
openai:gpt-4oFastest GPT-4-class model
openai:gpt-4o-miniLightweight and cost-efficient
openai:o3-miniStrong reasoning for complex problems
Automatic prefix caching: OpenAI automatically caches common prefixes server-side, so repeated context in long sessions is handled without any extra configuration.
Gemini models from Google DeepMind excel at long-context tasks and multimodal inputs.Get your key: aistudio.google.comRecommended models:
Model IDNotes
gemini:gemini-2.0-flashFast and capable for everyday tasks
gemini:gemini-2.5-proTop-tier reasoning and code quality
Automatic prefix caching: like OpenAI, Gemini handles prefix caching automatically on the server side.
GitHub Copilot integration uses OAuth — no API key is needed. Luminy authenticates using your GitHub account, so your Copilot subscription covers usage.How to connect:
1

Open Settings → GitHub Copilot

Navigate to the GitHub Copilot section in Luminy Settings.
2

Sign in with GitHub

Click Sign in with GitHub. A browser window opens for the OAuth authorization flow.
3

Authorize Luminy

Approve the authorization request. Luminy receives an OAuth token and stores it locally.
Model ID format: copilot:gpt-4o, copilot:claude-3.5-sonnet, etc. — Luminy fetches the list of models available to your Copilot subscription automatically after sign-in.
The ChatGPT provider connects to chatgpt.com via OAuth, giving access to models through your ChatGPT account rather than the OpenAI API.How to connect:Navigate to Settings → ChatGPT and click Sign in. Follow the OAuth flow in the browser. No API key is required — your ChatGPT account subscription is used.Model ID format: chatgpt:<model-id>
OpenRouter provides a single API key that routes to 100+ models from dozens of providers — a great option if you want to experiment with many models without managing multiple keys.Get your key: openrouter.ai/keysModel ID format: openrouter:anthropic/claude-3.5-sonnet, openrouter:google/gemini-flash-1.5, etc. — model IDs follow OpenRouter’s vendor/model naming convention.
OpenRouter is the fastest way to try a wide variety of models with a single API key. Many models also have free tiers available.
xAI’s Grok models are configured with a standard API key.Get your key: console.x.aiModel ID format: xai:grok-3, xai:grok-3-mini, etc.
Mistral AI offers the Mistral and Mixtral model families, known for strong performance at efficient parameter counts.Get your key: console.mistral.aiModel ID format: mistral:mistral-large-latest, mistral:codestral-latest, etc.
Groq’s LPU inference hardware delivers some of the fastest token generation speeds available in the cloud.Get your key: console.groq.comModel ID format: groq:llama-3.1-70b-versatile, groq:mixtral-8x7b-32768, etc.
Luminy also supports the following providers — all configured the same way (Settings → provider section → paste API key → Save):
ProviderGet your keyModel ID prefix
DeepInfradeepinfra.comdeepinfra:
Cerebrascloud.cerebras.aicerebras:
Coheredashboard.cohere.comcohere:
TogetherAIapi.together.xyztogetherai:
Perplexityperplexity.ai/settings/apiperplexity:
Alibaba DashScopedashscope.aliyun.comalibaba:
Venice AIvenice.aivenice:
Fireworksfireworks.aifireworks:
OpenCode Zen provides a set of free models that require no API key at all. They are always available in the model selector as soon as you launch Luminy.Available models:
Model IDDisplay name
opencode-zen:big-pickleBig Pickle
opencode-zen:mimo-v2.5-freeMiMo-V2.5 Free
opencode-zen:nemotron-3-ultra-freeNemotron 3 Ultra Free
opencode-zen:deepseek-v4-flash-freeDeepSeek V4 Flash Free
OpenCode Zen models are a great zero-cost option for quick tasks or for evaluating Luminy before committing to a paid API key.

​
Security note

Your API keys are stored locally on your device and used only for direct requests to the respective provider APIs. Luminy never proxies your keys through its own servers.