TinyHumans

TinyHumans Intelligence ships two products that work together: OpenHuman, a desktop application for personal AI agents that retains memory and preferences across sessions, and NeoCortex, a freemium API that provides the underlying memory infrastructure for production applications. The core workflow centers on persistence — agents remember what you worked on last Tuesday, adapt to your preferences over time, and execute multi-step tasks like email summarization, document organization, and calendar management without requiring the user to re-establish context each session. The agent layer supports autonomous execution: joining meetings, browser control, code execution, and scheduled tasks run without requiring you to manually trigger each step.

The differentiating feature is the memory architecture. Most agent frameworks treat memory as a session variable — it resets. NeoCortex treats memory as infrastructure: indexed, retrievable, and callable via API at $1/GB and $1/1M tokens on paid tiers, with a free tier for lower-volume use. For developers building production applications where context continuity matters — think support agents that remember a customer’s history, or research tools that accumulate a growing knowledge base — this is the layer that prevents context-from-scratch rebuilds on every call.

OpenHuman targets privacy-conscious users who are unwilling to route everything through a cloud AI vendor. The local-first framing is real: storage and retrieval run on-device. The caveat is that managed backend services — account authentication, model routing — still connect to TinyHumans-operated infrastructure. Teams with strict data residency or air-gap requirements need to audit exactly which calls leave the machine before committing. For users who are privacy-sensitive but not compliance-constrained, this is a meaningful step up from fully cloud-dependent alternatives. For teams under formal compliance regimes, the ‘local-first’ label requires scrutiny.