How to Use OpenHuman

Quick Start

After installing OpenHuman, the fastest path to value is: launch the app, verify the version, set your vault location, connect one low-stakes integration, and ask a cross-source question.

• Launch the app and confirm the version matches the release you downloaded.
• Set the Memory Tree vault path to a folder you back up.
• Connect one integration first — email or calendar are good starting points.
• Wait for the first sync to finish; background sync runs roughly every 20 minutes.
• Ask a question that pulls from the synced data to confirm memory is working.

First Launch Checklist

Before connecting any accounts, walk through these verification steps.

• Check that the app can reach the internet for OAuth and model routing.
• Verify the vault path is writable and excluded from aggressive cloud sync if you want it local-only.
• Review the permissions requested by the app: microphone, accessibility, and screen recording on macOS.
• Open Settings > Integrations and confirm the connector list loads.
• Do not connect primary accounts until you have completed the privacy checklist.

Using the Memory Tree

The Memory Tree is OpenHuman's core interface to your data. It organizes synced information into themes, entities, and documents so the assistant can answer questions without reloading full context.

• Themes are high-level buckets like work, family, or finance.
• Entities are people, companies, repos, or projects mentioned across sources.
• Documents are the original emails, notes, or transactions stored as Markdown.
• You can inspect and edit the vault in Obsidian or any Markdown editor.
• Changes you make in the vault are picked up on the next ingest cycle.

Local vs Cloud Model Routing

OpenHuman can route tasks to local models or cloud providers. Local AI keeps data on your machine but requires more setup.

• Enable local AI in Settings and point it to Ollama or LM Studio.
• Lightweight tasks like classification, formatting, and summarization route locally when enabled.
• Reasoning, coding, vision, and voice TTS route to cloud providers.
• Check the cost and privacy implications of each provider in the model routing settings.
• You can override routing per conversation if a specific task needs a different model.

Common Daily Workflows

These workflows show what OpenHuman is designed to do once memory is populated.

• Morning brief: ask for a summary of today's calendar, unread email priorities, and open GitHub mentions.
• Meeting prep: ask for context on attendees and recent threads related to the meeting topic.
• Research synthesis: ask OpenHuman to combine notes from Notion, Slack, and bookmarks on a topic.
• Voice commands: use Whisper STT to add quick notes or trigger actions without typing.
• Vault cleanup: delete stale chunks or refine entity mappings to improve answer quality.

Next Steps

Once you are comfortable with the basics, deepen your setup or compare OpenHuman to other tools.