#1515th of 18 in Memory
everos
Local-first agent memory runtime that stores memories as Markdown files
Documentation ↗
(opens in a new tab)Website ↗
(opens in a new tab)Repository on GitHub ↗
(opens in a new tab)
- Stars
- 13.4k
- License
- Apache-2.0
- Last commit
- Oct 2026
- Last release
- Sep 2026
- Language
- Python
Missed the latest GitHub check; these facts may be older.
Overview
EverOS is a Python library and HTTP server that stores agent conversations, files and trajectories as Markdown, then indexes them in SQLite and LanceDB for retrieval. It separates user memory (episodes, profile) from agent memory (cases, skills) and searches by user, agent, app, project and session IDs. A single OpenRouter key enables core add, flush and keyword search; embeddings and rerank unlock hybrid search, reflection and a Knowledge Wiki.
Who it is for: Developers adding persistent, inspectable memory to agents and coding assistants
Strengths
- Canonical Markdown files are editable and Git-versionable; a watcher re-syncs indexes after edits
- No MongoDB, Elasticsearch or Redis needed; storage is Markdown, SQLite and LanceDB
- Search can be scoped by user_id, agent_id, app_id, project_id and session_id
- Offline reflection merges episode clusters and refines profiles and skills between sessions
Weaknesses
- Needs an LLM provider (OpenRouter in the quick start); no fully offline setup is described
- Hybrid search, reflection and skill extraction require a separate embedding provider
- Knowledge Wiki and agentic search also need a rerank provider
- Office file ingestion needs LibreOffice installed on the host; no Docker setup is listed
What it needs
- no GPU
- Needs Python 3.12+, OpenRouter API key (or other LLM provider), LibreOffice (office documents only)
- Models: OpenRouter-hosted LLMs (default openai/gpt-4.1-mini), Optional embedding and rerank providers, Optional multimodal model (default google/gemini-3.8-flash)
- port 8000
Also in Memory
See all 18| Rank | Project | Score |
|---|---|---|
| 1 | gbrainLong-term memory store for coding agents, with sources and corrections | 85 out of 100 |
| 2 | hindsightAgent memory server with retain, recall and reflect operations | 80 out of 100 |
| 3 | CogneeMemory engine that turns documents and code into a knowledge graph | 79 out of 100 |
| #4 | MemPalaceLocal long-term memory store for agents with verbatim storage and semantic search | 78 out of 100 |
| #5 | agentmemoryPersistent cross-session memory for coding agents via hooks, MCP and REST | 77 out of 100 |