Back to Gap Map Product / UX / Agent memory & context stores

EverOS

EverMind
open source / Overall score: 2.8

EverOS is a local-first memory library and server for agents that writes conversations, files and agent trajectories to Markdown, extracts episodes, atomic facts, user profiles and reusable agent skills from them, and indexes the result in SQLite and LanceDB for retrieval. A background reflection pass merges related episodes and archives the fragments it replaces. EverMind builds it, and it was formerly called EverMemOS.

Openness

5 high confidence
5.0
license
Apache-2.0 (EverOS and the EverAlgo extraction, ranking and clustering packages it installs)
source
public
core features withheld
no — the 1,283-path tree at bc50821 has no ee/, enterprise/, commercial/ or proprietary/ path, and the eight everalgo packages pinned as dependencies are published from the public EverMind-AI/EverAlgo repository with source distributions
commercial
EverOS Cloud

EverOS and the EverAlgo libraries that hold its extraction and ranking algorithms are both published in full, and the memory runtime runs entirely on a user's machine. EverMind sells EverOS Cloud as a managed deployment of the same engine and memory format, not a version with more features.

Adoption

2 medium confidence
2.0

Adoption is read from monthly PyPI downloads of everos, the package that carries the library, server and command-line tool. EverOS Cloud and the separate MCP server package are not counted, and the evermemos package on PyPI is a client for the earlier EverMemOS hosted API rather than this code.

Capability

4 medium confidence
4.0

EverOS extracts several kinds of memory from raw conversations, consolidates them offline into merged narratives and rewritten profiles, and retrieves across separate user, agent and knowledge stores. Old memories leave search only when a newer narrative replaces them, never because of their age.

Verified 2026-11-20, except some scores, still awaiting re-confirmation