Your models, your machines
Local GGUF models run through a built-in llama.cpp, accelerated by Metal, Vulkan or CUDA. Cloud models (OpenAI, Anthropic, Gemini) are there when you choose them.
Local Cognitive
A private AI workspace for chats, agents and workflows. Models run on your own computer, or on your own GPU server that you connect with one command.
A GPU server in one command
Install the server on a Linux machine with NVIDIA GPUs, paste the key it prints into the app, and use it from anywhere. The connection is encrypted end to end.
Agents and workflows
Subagents, debates, a code mode with approvals, a visual workflow builder, tasks and schedules, MCP servers and plugins.
Synthesis
Describe a module’s contract, drive agents with a small flow language, and get code checked against that contract.
Start here
Section titled “Start here”Install on macOSApple silicon, signed and notarized.
Install on Windowsx64, Vulkan on any GPU, CUDA on NVIDIA.
First stepsDownload a model and start a chat.
Install the serverLinux x64 with systemd; NVIDIA optional.