OpenHuman
A free, open-source personal AI app for macOS, Windows and Linux, with local memory, multi-agent workflows, and local or cloud model backends.
OpenHuman is a free, open-source personal AI app for macOS, Windows and Linux that combines persistent local memory with agents that can act across your work apps. It's for people who want a desktop assistant that remembers context and handles tasks involving several tools. The project uses the GPL-3.0 license.
Its default memory engine stores information locally and mirrors it in an Obsidian vault. It decides what to retain, compress and retrieve when an agent needs context. You can also choose a remote memory service, including TinyHumans' hosted CortexDB, Mem0 or Supermemory. Model choices include Ollama, LM Studio, MLX and local OpenAI-compatible servers, alongside cloud providers such as OpenAI and Anthropic. The app and default memory run on your machine; cloud models and remote memory send the relevant data to those services.
Agents connect to Gmail, Notion, Slack, Telegram, GitHub and other work apps through integrations, MCP and skills. For repeatable tasks, you describe a workflow and the agent drafts a graph for you to review on a canvas. Workflows can include multiple agents, conditions, loops and approval steps, start on a schedule or app event, and resume after a pause.
A shared Rust core powers the desktop app, browser interface and terminal client. Developers can also embed that core in a Rust application. Multiple agents share one process, reducing the overhead of running a separate process for each agent, while each can have its own model provider, tools and sandbox.
