EverOS is a Python library and local-first memory runtime that gives agents and applications one portable long-term memory layer across coding assistants, apps, devices, and workflows. It stores conversations, files, and agent trajectories as Markdown, then syncs local SQLite and LanceDB indexes so the same material can be retrieved and reused.
Markdown files are the canonical store: they are readable, editable, diffable, and versionable with Git, and a cascade watcher re-syncs the indexes after a direct edit. Memory is split into a user track of episodes and profile and an agent track of cases and skills, and retrieval is scoped by userid, agentid, appid, projectid, and session_id. An offline reflection pass runs between sessions, merging episode clusters and refining profiles and skills.
It runs as a local HTTP server with memory endpoints under /api/v2 for add, flush, and search, plus an everos CLI for initialisation, server control, and a standalone demo of the ingest, extract, index, and recall loop. Python 3.12 or later and a single OpenRouter key cover the base setup; optional providers unlock the rest, and /health reports which capabilities are active.
| Configuration | Adds |
|---|---|
| [llm] only | Core memory flow and keyword search |
| Add [embedding] | Vector and hybrid search, reflection, and skill extraction |
| Add [rerank] too | Agentic search, agent hybrid search, and Knowledge Wiki |
| Add [multimodal] and the parser extra | Image, PDF, audio, and office-file ingestion |
Features
- Markdown source of truth: canonical .md files you can read, edit, diff, and version with Git, with a watcher that re-syncs the indexes
- Local three-part stack: Markdown plus SQLite and LanceDB indexes, without a separate database service
- Two memory tracks: user episodes and profile alongside agent cases and skills, each a first-class surface
- Orthogonal retrieval: search scoped by user, agent, app, project, and session identifiers
- Knowledge Wiki: editable, source-backed Markdown knowledge pages with taxonomy, CRUD APIs, and topic search
- Reflection: offline consolidation that merges episode clusters and refines profiles and skills between sessions
- Multimodal ingestion: image, PDF, audio, HTML, email, and office files through an optional extra; office formats require LibreOffice on the host
- Server and CLI: local server with /api/v2 memory endpoints, capability reporting on /health, and a demo command that needs no key
- Agent integrations: plugins for OpenClaw, Hermes, Deepseek Harness, and Dify, and it ships built into Raven
Integrates with
Claude Code
Terminal-based agentic coding tool that reads your codebase and runs tasks through natural-language commands
DeepSeek Harness
DeepSeek open-sourced its full agent harness built so every capability is a plugin. You can swap the loop, the sandbox,
Dify
Open-source LLM app development platform combining a visual workflow canvas, RAG pipelines, agents, model management, and observability
Hermes
Self-improving AI agent with a learning loop that creates and refines skills, recalls past sessions, and runs across chat platforms