Verified project record
tinyhumansai/openhuman
OpenHuman is a local-first personal AI application for persistent memory, durable multi-agent workflows, and research across personal data and the web. It combines local memory storage with optional external services, routed models, integrations, and an enforced Privacy Mode.
Project overview
It combines editable local Markdown memory mirrored to Obsidian with checkpointed agent graphs that can pause, survive restarts, and resume mid-run.
- Project type
- AI Search · Model Runtime · Prompt Engineering
- Use cases
- Chat Assistants · Knowledge Q&A · Search & Research
- Deployment
- Refer to project documentation
- License
- GPL-3.0
Best for
- General users who want persistent personal memory represented in editable Markdown and mirrored to Obsidian.
- Researchers and developers who need web research, tool use, routed models, and durable agent workflows with human pauses and resumptions.
- Users who can choose between a managed model route, a user-provided provider, or a fully local Ollama model according to their data-boundary needs.
Key capabilities
- Compresses user data into scored Markdown memory trees stored in local SQLite and mirrored to an editable Obsidian vault.
- Uses documented account integrations and periodic automatic fetching to feed connected data into the memory system.
- Maintains long-term goals, durable goals for individual threads, and a shared kanban board for each conversation.
- Runs checkpointed agent graphs, delegates work to specialist agents, survives restarts, and supports human pauses and resumptions.
- Provides documented web search, scraping, coding tools, browser operation, native voice processing, and workload-specific model routing.
- Provides documented image and video generation in the user's workspace through named external generation models.
- Provides an enforced mode in which inference does not leave the user's machine, alongside encrypted local data and operating-system keyring secret storage.
- Compresses tool output before it reaches the model to reduce token consumption while retaining the same information.
Limitations and risks
- The application is an early beta under active development, so users should expect rough edges that may affect stability or usability.
- Outside Privacy Mode, documented search, provider, account, and messaging integrations may involve external services or accounts. Local orchestration or local execution alone does not establish an end-to-end local-only or offline data boundary.
- GPU requirements, minimum hardware, telemetry behavior, and whether any supported installation targets require an OpenHuman subscription are not documented in the supplied project facts.
Getting started
- Download an installer from the project website or GitHub Releases, then follow the UI-first installation path. The documented user path requires no configuration files or terminal use and is described as reaching a working agent in a few clicks.
Evidence and sources
- README: <strong>OpenHuman is your personal AI super intelligence: a brain that remembers everything, a fantastic orchestrator, a deep researcher. Local-first, simple, powerful.</strong>
- README: OpenHuman is three things most assistants aren't: **a brain** that builds a persistent, local memory of your world; **a fantastic orchestrator** that runs fleets of agents on dura…
- README: > **Early Beta**: Under active development. Expect rough edges.
- README: - **[Memory Tree](https://tinyhumans.gitbook.io/openhuman/features/memory-tree) + [Obsidian Wiki](https://tinyhumans.gitbook.io/openhuman/features/obsidian-wiki)**: your data comp…
- README: - **[100+ OAuth integrations, 5,000+ MCP servers, 90,000+ Skills](https://tinyhumans.gitbook.io/openhuman/features/integrations)**: one click into Gmail, Notion, GitHub, Slack and…
AI Search
Find projects, verify facts, compare options, or turn a complex need into an actionable plan
Try a searchA click only fills the search box; you stay in control
Project Details
0