Tool not found.
Back to Home
Open-source AI agent that runs entirely on local models, fully private and offline, with hardware profiling and no cloud dependency, API keys, or subscriptions required.
Magnitude is an open-source AI agent with local models built directly in, designed to run fully private and offline on any hardware. Rather than requiring users to separately set up Ollama, a model server, or inference infrastructure, Magnitude profiles the user's hardware and automatically recommends and downloads the best-fitting models, then configures and runs them itself. Everything, including prompts, files, and the models themselves, stays entirely on the user's machine with nothing sent to the cloud, and it works completely offline once set up. The agent can be extended with reusable 'skills' — capabilities like driving a logged-in Chrome browser, or reading and building Excel, PowerPoint, Word, and PDF files — installable via the skills.sh directory. Users aren't limited to Magnitude's built-in model catalog either: compatible GGUF models from Hugging Face can be downloaded and used directly, and an OpenAI-compatible inference endpoint can be connected for custom setups. Since there are no token costs, API keys, or rate limits, Magnitude is free to run indefinitely. The project is Apache 2.0 licensed, fully inspectable, and supports macOS and Linux natively, with Windows supported through WSL.
