openhuman

Why we included this project

OpenHuman is a desktop-first agent platform that pairs a persistent local memory with orchestration and deep research. Instead of starting each session cold, it keeps a running record of your work: data is compressed into editable Markdown trees in SQLite, mirrored as an Obsidian vault you can open and edit, and refreshed by a background loop. The orchestration side proposes approval-gated workflows on a visual canvas and runs checkpointed agent graphs you can steer when they stall and replay with per-call costs. Its research layer sweeps your memory and files as you type, and native tools cover web search, a real browser, voice, and meeting agents for Meet, Zoom, Teams, and Webex. Teams and power users who want a local-first, human-in-the-loop setup rather than a thin chatbot wrapper will find model routing that mixes a subscription, your own provider keys, or fully local Ollama models.