Your models.
Your keys.
Commandr runs on five cloud providers with your own API key, or two that run free on your Mac. Give each agent the model that suits its job.
Seven providers, one app.
Every provider goes through the same chat API, so every agent keeps its file tools whichever you pick.
In the cloud · with your API key
- DeepSeekChat and reasoning models at a low price.Text onlyGet a key ↗ from DeepSeek
- OpenAIGPT and o-series models.Reads imagesGet a key ↗ from OpenAI
- AnthropicClaude models.Reads imagesGet a key ↗ from Anthropic
- Google GeminiGemini models from Google AI Studio.Reads imagesGet a key ↗ from Google Gemini
- OpenRouterOne key for models from many providers.Most read imagesGet a key ↗ from OpenRouter
On your Mac · free
- OllamaStart Ollama and its models show up by themselves.Get Ollama ↗
- LM StudioStart LM Studio’s local server and its models show up by themselves.Get LM Studio ↗
A model for each agent.
Put your strongest model on Architect and a cheaper one on Document. Choose in Settings → Models, or change one agent for a single project in Project settings → Agents.
- Each agent runs on the provider of the model you pick, so one project can mix them.
- Visual checks need a model that can see images. Automatic picks one for you.
- Images sent to a text-only model are left out, and the reply says so.
Connected in a minute.
Get a key
Create an API key in your provider’s console — or install Ollama or LM Studio and download a model.
Paste it into Commandr
Open Settings with ⌘, and go to AI providers. Your key is encrypted on your Mac and only sent to that provider.
Pick your models
In Settings → Models, set a default and give any agent its own. Change them any time.
Run it all on your Mac.
With Ollama or LM Studio, agents run on local models: no API bills, and nothing from your projects sent to a provider. Commandr still needs an internet connection to check your account.
- $0per run
- Nothingsent to a provider
- Mix local and cloud models in one project
Bring a key. Or don’t.
Commandr is free during the beta. You only pay your provider for what the agents use — see pricing.