OpenHuman by TinyHumans

What is OpenHuman by TinyHumans?

OpenHuman is TinyHumans' flagship product: a local-first, open-source AI agent (built on Rust + Tauri, licensed under GNU GPL3) designed to be your persistent memory and autonomous executor across every tool you use. Rather than a stateless chatbot, OpenHuman ingests your emails, messages, commits, notes, and calendar through 118+ one-click OAuth integrations, stores everything in a deterministic Memory Tree on your own machine, and keeps context continuously updated via auto-fetch every 20 minutes. Powered by NeoCortex — TinyHumans' proprietary memory layer — it processes up to 10 million tokens at 4,000 tokens/second, building a living knowledge graph of your world. Smart token compression (TokenJuice) and automatic model routing keep costs low while routing tasks to the right model. A desktop mascot gives OpenHuman a face: it speaks, lip-syncs, joins Google Meet calls, and thinks in the background even when you're not typing.

Key Features

• 1B-Token Persistent Memory

  Stores up to 1 billion tokens of personal context locally in SQLite via a deterministic Memory Tree — chunked, scored, and folded into per-source/per-topic/per-day summaries you can inspect as Markdown files in an Obsidian-style vault.

• 118+ One-Click Integrations

  Native OAuth connections to Gmail, Slack, GitHub, Notion, Google Drive, Linear, Jira, Stripe, Calendar, and many more — no API keys or plugin setup required.

• Auto-Fetch & Always-On Context

  Pulls fresh data from every active connection every 20 minutes, keeping the agent's understanding of your world current without any manual prompting.

• NeoCortex Memory Engine

  Processes 10M+ tokens at up to 4,000 tokens/second using GraphRAG and noise-pruning to build a typed knowledge graph that sharpens over time rather than accumulating noise.

• Full Agent Toolbelt

  Built-in tools include web search, web scraping, a complete coder toolset (filesystem, git, lint, test), browser and computer control, cron scheduling, sub-agent coordination, and native voice with STT/TTS.

• Automatic Model Routing & Local AI

  Routes each task to the best-fit model (frontier for reasoning, fast/cheap for lightweight tasks, vision for visual input) under one subscription; optional local inference via Ollama or LM Studio keeps sensitive workloads on-device.

Use Cases

• Personal Knowledge Management : Professionals and researchers can connect all their tools and let OpenHuman maintain a living, searchable map of everything they know, have written, or have discussed.
• Email & Communication Automation : Users can instruct OpenHuman to read, summarize, draft, and send emails or messages across connected accounts without switching between apps.
• Developer Workflow Assistance : Engineers can give OpenHuman access to their repos, issues, and PRs; the agent can search code, run tests, and surface relevant context from past commits or Slack threads.
• Meeting Intelligence : OpenHuman joins Google Meet calls as a live participant, transcribes them directly into the Memory Tree, and can respond verbally during the meeting.
• Long-Horizon Task Execution : By combining persistent memory, scheduling (cron), and sub-agent coordination, OpenHuman can autonomously manage multi-step workflows that run over days or weeks.

FAQs