跳到主要内容

Getting Started

This page walks you through installing OpenHuman, going through the in-app onboarding, and running your first request.

OpenHuman is open source under the GNU GPL3 license. The codebase is at github.com/tinyhumansai/openhuman.

System Requirements​

OpenHuman runs on macOS, Windows and Linux desktops. 4 GB+ RAM is recommended; 16 GB+ if you intend to ingest very large mailboxes or repos, or run a local model on the same machine.

Permissions​

The first time you launch OpenHuman, the OS will prompt for the permissions the app needs (Accessibility on macOS, Input Monitoring for the voice hotkey, Camera/Microphone if you plan to use the Meeting Agent). You can review and adjust these any time under Settings → Automation & Channels.

1. Download and Install​

Get the OpenHuman desktop app from tinyhumans.ai/openhuman or via your platform's package manager. Open the app once it's installed.

2. Sign In​

The first screen is "Sign in! Let's Cook". Multiple sign-in options are available, including social login. There's also an Advanced panel for pointing the app at a custom core RPC URL if you're running your own backend; most users can ignore it.

No permanent lock-in. Signing in does not grant OpenHuman ongoing access to anything. All third-party access requires explicit OAuth approval per integration in the steps below.

Know what is local and what is managed. Your Memory Tree database, Markdown vault, workspace config, and local runtime state live on your machine. The default setup still uses OpenHuman-hosted services for sign-in, model routing, managed integration OAuth/tool calls, and web search proxying.

3. Run Your First Request​

Once Gmail has been ingested (the first auto-fetch tick happens within twenty minutes), try prompts like:

Briefings

  • "What do I need to know from the last 12 hours?"
  • "What's waiting on me?"

Cross-source queries

  • "Summarize what I missed today."
  • "What are the key decisions from this week?"
  • "Extract action items from my recent conversations."
  • "What did Sarah say about the project across email and chat?"

OpenHuman picks the right model for each task automatically. See Automatic Model Routing.

4. Open the Obsidian Vault​

The Memory tab has a View vault in Obsidian button. Click it to open <workspace>/wiki/ in Obsidian. You can browse the agent's summaries, drop in your own notes, and even build manual links - the agent will pick up your edits on the next ingest.

5. Let the Mascot Do More​

Now that the agent has memory and a model, the rest of the product is about giving it more surfaces:

  • Meeting Agents - drop a Google Meet link in and the mascot joins as a real participant: it listens, takes notes into the Memory Tree, speaks back into the call, and uses tools live.
  • Auto-fetch from Integrations - connect more sources from Settings; every twenty minutes the scheduler pulls fresh data into your tree.
  • Native Voice - push-to-talk dictation and TTS replies so you can talk to OpenHuman instead of typing.
  • Subconscious Loop - let the mascot keep working on standing tasks while you're away.

Next Steps​