EverOS

EverOS is a Python library and local-first runtime that adds durable memory to AI agents across platforms and workflows. It stores conversations, files, and agent execution traces as readable Markdown files, then indexes them with SQLite and LanceDB for fast retrieval. The system automatically extracts and categorizes memories as profiles, episodes, or skills, and offline consolidation turns repeated successful task patterns into reusable skills shared across agent teams.

Designed for developers building multi-agent systems, EverOS integrates with Claude Code, Codex, OpenClaw, Hermes, and MCP-compatible tools without requiring infrastructure changes. It achieves 93%+ retrieval accuracy with sub-500ms latency while reducing token usage by roughly 10×. All data remains portable and version-controllable in Markdown format, deployable on EverOS Cloud or self-hosted under Apache 2.0.