What website is this?
TinyHumans is an open-source AI agent suite built around people, with OpenHuman (a desktop personal agent) and NeoCortex (production-oriented AI memory infrastructure) presented on the same site. It focuses on turning email, calendars, docs, and code-hosting context into durable memory your agent can reuse—more of an “agent harness + long-term memory” stack than a one-off chat page. OpenHuman is still in Early Beta; NeoCortex API access has often been in closed testing—check the site and docs for current availability.
Key Features
- OpenHuman: Desktop agent that connects Gmail, Notion, GitHub, Slack, and more via OAuth and exposes them as typed tools.
- Memory Tree + Obsidian Wiki: Local SQLite storage with Markdown knowledge organized for browsing and editing.
- Scheduled auto-fetch: Periodically pulls updates from connected accounts (docs cite ~20 minutes) into the local memory tree.
- NeoCortex memory API: Ingest and recall memory, batch document import, plus Python and other SDKs and agent-framework plugins.
- Built-in toolset: Web search, scraping, coding tools (filesystem, git, tests), and voice I/O per official docs.
- Token compression (TokenJuice): Compresses tool output before it hits the LLM to control token use on the OpenHuman side.
Use Cases
- Indie developers link GitHub and calendar so the desktop agent summarizes recent issues, PRs, and meetings before they start coding.
- Knowledge workers connect Notion and Gmail to sync mail and docs into a local Obsidian-style vault for themed retrieval and drafting.
- Product teams embed NeoCortex in their own agents, using namespaces for preferences, session memory, and document chunks across turns.
- Users who want a persistent desktop presence use the mascot, voice, and meeting-related features (scope per GitBook documentation).
- Teams already on backends like agentmemory can switch OpenHuman’s memory backend per docs for compatibility.
Who is it for?
- Individuals and small teams who want local-first context, OAuth-linked workflows, and durable agent memory.
- Developers evaluating an API-callable memory layer (write, recall, document management) for LLM products.
- Technical users comfortable with desktop install, API keys, and following open-source repos and docs.
- May not fit: People who only want occasional browser chat and won’t install apps or authorize third-party accounts.
- May not fit: Users who need one-shot Q&A only, don’t care about memory decay/sync, and don’t want Beta instability.
How It Compares to Similar Tools?
Two common patterns are terminal/plugin agent harnesses and standalone vector memory or RAG services. OpenHuman skews UI-first desktop agent: built-in memory tree, broad OAuth, and auto-fetch to shorten cold-start “the agent doesn’t know me yet” time. NeoCortex targets production memory infrastructure (interaction signals, time decay, conscious recall) for embedding in your own agents—not replacing full office suites.
If you need a zero-install, browser-only lightweight assistant, OpenHuman’s path may feel heavy; if you care more about calling a memory API inside your service, weigh NeoCortex first (access per current site notices). Neither replaces a professional IDE, PM tool, or mail client—they sit above as agent and memory layers.
FAQs
Q: Is TinyHumans free? How does open source relate to pricing?
A: Repos are open source (license per README), but hosted APIs, model routing, and subscriptions follow the site and pricing pages—no specific prices asserted here.
Q: Do I have to install OpenHuman to use NeoCortex?
A: No. NeoCortex is mainly API/SDK integration; OpenHuman is a separate desktop product and can be used alone or together.
Q: Can anyone call the NeoCortex API right now?
A: Not necessarily. Official README has described closed testing; whether it’s open and how to get keys depends on tinyhumans.ai and docs today.
Q: Who is TinyHumans a good fit for?
A: Technical users willing to give agents workflow context and handle Beta plus local/API setup; casual Q&A without account linking is usually a weaker match.
