OpenHuman

What is OpenHuman?

OpenHuman is an open-source desktop assistant designed to remember a user's context and take actions across connected tools. It builds a local-first Memory Tree from documents, email, chats, notes, and application data, then stores the underlying knowledge as an Obsidian-compatible Markdown vault that users can inspect and edit.

The product includes integrations, web research, coding and file tools, scheduling, browser and computer control, voice features, and model routing. Its privacy model is hybrid rather than fully offline by default: memory, the Markdown vault, configuration, and local runtime state stay on the device, while managed sign-in, model routing, web search, and some OAuth integrations use hosted services unless users configure alternatives. OpenHuman is in early beta, so users should expect rough edges and avoid unsupervised high-impact actions.

How to Use OpenHuman

1. Download and install the desktop application for a supported operating system.
2. Complete onboarding and choose managed services, custom credentials, or available local-model options.
3. Connect only the accounts and tools the assistant needs.
4. Let the application import and organize selected information into its local memory and Markdown vault.
5. Start with low-risk questions and actions, then inspect memory and permissions before enabling broader automation.

Core Features

• Local-first Memory Tree: Organizes connected information into hierarchical summaries stored on the device.
• Editable knowledge vault: Writes source chunks to an Obsidian-compatible Markdown collection.
• Application integrations: Connects services such as email, calendars, repositories, notes, project tools, and messaging.
• Agent toolset: Includes web, file, coding, scheduling, browser, computer-control, and coordination tools.
• Model routing and local AI: Routes tasks across models and supports optional local workloads through compatible runtimes.
• Voice and meeting features: Supports speech input and output and related meeting-assistant workflows.

Use Cases

• Personal knowledge memory: Build a searchable, inspectable record across notes, messages, documents, and connected accounts.
• Daily coordination: Ask the assistant to summarize context, prepare work, or coordinate recurring personal workflows.
• Cross-application automation: Perform instructed tasks across email, calendars, files, repositories, and project tools.
• Research and coding: Combine web retrieval, local files, terminal tools, and persistent context in one desktop assistant.

Pricing

OpenHuman's desktop code is available under the GNU GPLv3 license. The product page describes one subscription for managed access to multiple AI providers, but it does not publish a reliable current price in the page content reviewed here. Users can configure some custom or local providers, although certain real-time triggers and hosted features still require managed services. Review the current in-app plan and provider costs before enabling paid usage.

Frequently Asked Questions

Is all OpenHuman processing local?

No. Memory, the Markdown vault, configuration, and local runtime state are stored on the device, but the default setup uses hosted services for functions including sign-in, model routing, web search, and managed integrations.

Is OpenHuman open source?

Yes. Its public desktop code is licensed under GNU GPLv3. Some hosted services remain separate from the local client.

Is OpenHuman production ready?

The project describes itself as an early beta under active development. Users should expect instability, restrict permissions, and supervise consequential actions.