Local-first · macOS · open source
Meet Cosmo
A tiny companion that lives on your desktop — big expressive eyes, a voice you can talk to, and a knack for actually getting things done. Childlike heart, sharp mind.
Tip: move your cursor — he's watching. 👀
Plug in any brain
OpenAI
Anthropic
Gemini
Grok
DeepSeek
Groq
Cerebras
Ollama · fully local
👀
Expressive, alive
Big animated eyes that blink, glance around, follow your cursor, and shift mood — curious, thinking, happy, sleepy. He feels present, not like a chat box.
🎙️
Just talk to him
On-device speech recognition, a wake word ("Cosmo"), natural end-of-turn detection, and a warm voice back. Push-to-talk by default; go hands-free if you like.
🧰
Actually does things
Searches the web, opens apps and sites, sets timers and reminders, takes notes, checks the weather, controls music — through a clean, extensible tool system.
Yours, and only yours
Private by architecture, not by promise
Cosmo is built local-first. The things that should stay on your machine, stay on your machine — it's wired into how he's built.
No keylogging
He senses whether you're working from idle-time only — never key codes, never what you type.
No camera, ever
Cosmo senses focus from idle-time and your active app — locally. It never opens your webcam, so there's nothing to leak.
Your brain, your keys
Run a local model with Ollama, or bring your own API key for a frontier model. You choose.
One buddy, many brains
Swap the model, keep the character
Cosmo speaks to OpenAI, Anthropic, Google Gemini, xAI Grok, DeepSeek, Groq, Cerebras — or a fully local model via Ollama. The personality, the eyes, and the tools stay exactly the same.
Free to use. A coffee keeps the eyes blinking. ☕
Cosmo is open source. If he makes you smile, you can chip in — totally optional.
Questions
Cosmo FAQ
What a local-first, private AI desktop assistant actually means — in plain terms.
What is Cosmo?
Cosmo is a free, open-source AI desktop assistant for macOS — a small, always-on-top animated character you can talk to. It has an on-device voice pipeline (speech-to-text and text-to-speech) and a pluggable LLM brain that runs locally or in the cloud.
Does Cosmo run locally and work offline?
Yes. Cosmo is local-first: speech recognition and the "Cosmo" wake word run on-device, and you can run the language model fully locally with Ollama. With a local model, Cosmo needs no internet connection and no cloud account.
Is Cosmo private? Does it send my data to the cloud?
Privacy is built into the architecture. Cosmo never opens your webcam, never logs keystrokes, and never sends your window titles or screen anywhere. The AI only ever sees what you explicitly type or say — and with a local model, nothing leaves your machine.
What is the difference between local and cloud TTS and STT?
Speech-to-text (STT) turns your voice into text; text-to-speech (TTS) turns Cosmo's replies into a spoken voice. Cosmo runs STT on-device and offers both a local voice and cloud voices. Cloud voices can sound richer; local keeps everything offline and free.
Is Cosmo free?
Yes. Cosmo is free and open source under the Apache-2.0 license. You only pay if you choose a cloud LLM or voice and bring your own API key; with Ollama and the local voice it is completely free.
Which AI models and providers does Cosmo support?
Cosmo's brain is pluggable: OpenAI, Anthropic, Google Gemini, xAI Grok, DeepSeek, Groq and Cerebras, or a fully local model via Ollama. The character, eyes and tools stay the same whichever you choose.