EverMind Launches EverOS Public Beta with Self-Evolving Memory for AI Agents

EverMind has launched the global public beta of EverOS, a memory layer built for self-evolving AI agents, announced in a press release. The company also revealed a brand upgrade as part of its effort to enable agents to learn from experience and adapt autonomously.

EverOS introduces a Skills Evolution Engine that allows agents to extract and cluster task data into reusable skills. According to internal benchmarks, this approach improved success rates in software engineering tasks by over 230% when compared to baseline models. The platform’s multimodal retrieval system (mRAG) supports documents, images, and URLs through a single API, combining dense semantic and sparse keyword retrieval for more accurate cross-modal recall.

The system’s HyperMem architecture, based on a paper accepted at ACL 2026, replaces traditional vector databases with a hypergraph memory network, enabling faster and more complex associative reasoning. EverOS is available for developers through a RESTful API and features integrated coding and chat playgrounds for testing and comparison.

During the beta phase, EverMind is launching a developer co-creation program to encourage plugin and use-case development. Participants can register on the EverOS website or GitHub to contribute and test the system.