Agent memory的创业方向好火爆！ 很多团队都在做，大概有下面几种比较主流的思路： 第一种最粗暴：context路线，把上下文增长。 第二种最常见：RAG / 向量库路线，接一个向量库把历史内容 embedding 后做检索。 第三种是产品思维：memory API路线，把记忆封装成 API，直接调用。 三种方向，其实利弊都很突出！

我越来越认可的方向，是EverMind团队最近在GitHub开源的项目：EverOS 非常值得关注，早早就点上了star！ GitHub repo: https://github.com/EverMind-AI/EverOS… 他们提出了第四种思路，解决方案是： 把记忆做成一个可读、可编辑、可版本管理、可被人和 Agent 共同维护的知识层。 不是再做一个 Agent，而是在做 Agent 真正缺的一层基础设施：memory layer。

我觉得EverOS有几个特别务实，而且非常亮眼的特点： 第一，是把 Markdown 作为记忆源。 记忆是以 Markdown 保存，你可以直接打开、阅读、编辑、管理，甚至放进 Obsidian 里，人和agent都可以读，这是可检查的！ 第二，两条记忆轨道并行 用户记忆（user memory）回答“这个人是谁、偏好什么、过去发生过什么”， Agent 记忆（agent memory）回答“我做过哪些任务、哪些流程有效、哪些错误不要再犯”。 这个设计很重要！这两种记忆区分开，后面做权限、检索、迁移、复用都会非常的顺畅，并且几乎不会出幺蛾子。 第三，非常务实的本地栈 本地三件套“Markdown + SQLite + LanceDB”，既能本地跑，又能支持结构化过滤、BM25 和向量检索。 第四，多模态摄入 这比只支持 chat history 实用很多，因为用户的上下文从来不只存在于聊天框里，它可能在文本、图片、网页、会议录音、邮件和代码仓库里，不同格式内容都可以统一进 searchable memory。 第五，自演化能力 这点我最喜欢，很多团队做成了单独产品！ EverOS 不是简单把记忆“存一下，再搜出来”，它还能从真实使用中提取常见 cases 和 skills，让重复出现的工作模式变成可复用能力，而不是每次都让 Agent 重新摸索。

EverOS这个整体方案，感觉会成为下一代Agent的落地方向： 记忆不只是 recall，而应该让系统逐渐更懂任务、更懂用户、更懂自己。

当然，Agent memory 仍然有很多难题： 比如什么该记、什么不该记、如何避免错误记忆、权限隔离、记忆质量衡量、处理隐私和同步等等。 这些，都还需要更优的解决方案，也正因为这些问题还没完全解决，开源项目才值得更多 builder 参与、试用、提 issue、做 demo。 所以，有能力的开发者大佬，不妨持续关注一下。 这个世界，需要更多的 builder ~

EverMind-AI/EverOS

Source: https://github.com/EverMind-AI/EverOS

EverOS 1.0.0

EverOS 1.0.0 is a major release for self-evolving memory. It brings a local-first runtime, Markdown as the source of truth, hybrid retrieval, multimodal ingestion, user and agent memory scopes, and modular algorithms through EverAlgo.

Coming next: Knowledge Wiki will turn memory into editable, source-backed Markdown knowledge pages. Reflection will run when the system is idle or offline to connect signals, compress history, and improve profiles and skills between sessions.

Why Ever OS

EverOS is the local memory operating system for agents and makers. It gives one portable memory layer across coding assistants, apps, devices, and workflows. Today it stores conversations, files, and agent trajectories as readable Markdown, then syncs local SQLite and LanceDB indexes for fast retrieval and self-evolving reuse.

Quick Start

Goal: start EverOS, write one memory, and search it back.

0. Prerequisites

• Python 3.12+
• API keys for the default providers: OpenRouter for chat / multimodal, and DeepInfra for embedding / rerank. You can use other OpenAI-compatible providers by changing the matching *__BASE_URL fields in .env.

1. Install

uv pip install everos
# or: pip install everos

2. Configure

Generate a starter .env file, then fill the four API key slots shown in the generated comments. Only two distinct keys are needed with the defaults: OpenRouter for LLM / MULTIMODAL, and DeepInfra for EMBEDDING / RERANK.

everos init
# or, from a source checkout:
cp .env.example .env

everos init writes ./.env by default. Use everos init --xdg to write ${XDG_CONFIG_HOME:-~/.config}/everos/.env instead.

3. Start EverOS

everos server start

Keep the server running, then open a second terminal and check it:

curl http://127.0.0.1:8000/health

Expected response:

{"status":"ok"}

everos server start searches for .env in this order: --env-file <path> → ./.env (cwd) → ${XDG_CONFIG_HOME:-~/.config}/everos/.env → ~/.everos/.env. The endpoint stack is OpenAI-protocol compatible (OpenAI / OpenRouter / vLLM / Ollama / DeepInfra) - override *__BASE_URL in the generated .env to point at any of them.

4. Try Your First Memory

Add a tiny conversation:

TS=$(($(date +%s)*1000))

curl -X POST http://127.0.0.1:8000/api/v1/memory/add \
  -H 'Content-Type: application/json' \
  -d "{
    \"session_id\": \"demo-001\",
    \"app_id\": \"default\",
    \"project_id\": \"default\",
    \"messages\": [
      {\"sender_id\": \"alice\", \"role\": \"user\", \"timestamp\": $TS, \"content\": \"I love climbing in Yosemite every spring.\"},
      {\"sender_id\": \"alice\", \"role\": \"user\", \"timestamp\": $((TS+10000)), \"content\": \"My favorite coffee shop is Blue Bottle in SOMA.\"}
    ]
  }"

Force extraction for the local demo:

curl -X POST http://127.0.0.1:8000/api/v1/memory/flush \
  -H 'Content-Type: application/json' \
  -d '{"session_id":"demo-001","app_id":"default","project_id":"default"}'

Search it back:

curl -X POST http://127.0.0.1:8000/api/v1/memory/search \
  -H 'Content-Type: application/json' \
  -d '{
    "user_id": "alice",
    "app_id": "default",
    "project_id": "default",
    "query": "Where do I like to climb?",
    "top_k": 5
  }'

You should see the Yosemite memory in the response. If the result is empty on the first try, wait a moment and retry; Markdown is written synchronously, while the local index catches up in the background.

First memory unlocked. You just gave EverOS a fact, flushed it into durable Markdown-backed memory, and searched it back through the local index. That is the core loop. Want to see the source of truth? Open ~/.everos and inspect the generated Markdown files.

For annotated responses and the Markdown files EverOS creates, see QUICKSTART.md.

Optional: Ingest Multimodal Files

To ingest non-text content (image / pdf / audio / office documents) through /api/v1/memory/add content items, install the optional extra:

uv pip install 'everos[multimodal]'   # or: pip install 'everos[multimodal]'

This pulls in everalgo-parser (with the [svg] bundle for SVG support via cairosvg) and wires up the multimodal LLM client (EVEROS_MULTIMODAL__* fields in .env, defaults to google/gemini-3-flash-preview via OpenRouter).

Office document support requires LibreOffice as a system dependency. The parser shells out to soffice (LibreOffice’s headless renderer) to convert .doc / .docx / .ppt / .pptx / .xls / .xlsx to PDF before feeding the result into the multimodal LLM. Without LibreOffice, office uploads return HTTP 415 with a clear error message; PDF / image / audio / HTML / email parsing is unaffected.

Install on the host before serving office documents:

brew install --cask libreoffice              # macOS
sudo apt-get install -y libreoffice          # Debian / Ubuntu

For Contributors

git clone https://github.com/EverMind-AI/EverOS.git
cd EverOS
uv sync                              # creates ./.venv and installs deps
source .venv/bin/activate            # or prefix commands with `uv run`
everos init                          # fill the four API key slots in .env (two distinct keys)

everos --help
make test

Use Cases

Now that you have had your first successful EverOS moment, explore what people are building with persistent memory across agents, apps, and community integrations.

Use cases show what persistent memory makes possible in real products and workflows. Some examples are packaged in this repository; others point to external demos or integrations you can study and adapt.

Reunite - Find With EverOS Parents describe what they remember. Children describe what they recall. Reunite uses semantic memory to surface the connections. Learn more	Hive Orchestrator Browser-native hive-mind for CLI coding agents - Claude Code, Codex, Gemini, and OpenCode collaborate as real PTY processes via a team protocol. Code
AI Coding Assistants With EverOS Universal long-term memory layer for AI coding assistants, powered by EverOS. Code	AI Data Technician An agentic AI system that learns from scientist interaction to inspect, analyze, and classify high-dimensional time series data - with persistent memory that improves across sessions. Code
Rokid AI Assistant With EverOS Connect to EverOS within Rokid Glasses enabling long-term memory for all of your smart activities. Coming soon	Creative Assistant With Memory Creative assistant with long-term memory, so your creative context stays available across sessions. Coming soon
Earth Online Memory Game Earth Online is a memory-aware productivity game that turns everyday planning into a living quest log. Code	Multi-Agent Orchestration Platform Golutra presents a multi-agent workforce for engineering teams, extending the IDE model from a single assistant to coordinated agents. Code
Your Personal Tasting Universe Record, visualize, and explore your tasting journey through an immersive 3D star map. Code	EverOS Open Her Build AI that feels. Open-source persona engine - personality emerges from neural drives, not prompts. Inspired by Her. Code
Browser Agent For Personal Memory Ruminer brings persistent memory to a browser agent so it can carry personal context across web tasks. Plugin	EverMem Sync With EverOS One command to connect any AI coding CLI to EverMemOS long-term memory. Code
MCO - Orchestrate AI Coding Agents MCO equips your primary agent with an agent team that can work together to solve complex tasks. Code	Study Buddy With Self-Evolving Memory Study proactively with an agent that has self-evolving memory. Code
Alzheimer’s Memory Assistant Empowering individuals with advanced memory support and daily assistance. Code	Memory-Driven Multi-Agent NPC Experience An iOS sci-fi mystery game where players explore and uncover the truth. Code
Mobi Companion An iOS app where users create, nurture, and live with a personalized AI companion called Mobi. Code	AI Wearable With Memory A context-native AI wearable that listens to everyday life and converts conversations into memory. Code
Legacy OpenClaw Agent Memory Archived pre-1.0.0 plugin reference. New integrations should use the EverOS 1.0.0 API. Learn more	Live2D Character With Memory Add long-term memory to a real-time Live2D character, powered by TEN Framework. Code
Computer-Use With Memory Run screenshot-based analysis with computer-use and store the results in memory. Live Demo	Game Of Thrones Memories A demonstration of AI memory infrastructure through an interactive Q&A experience with A Game of Thrones. Code
Claude Code Plugin Persistent memory for Claude Code. Automatically saves and recalls context from past coding sessions. Code	Memory Graph Visualization Explore stored entities and relationships in a graph interface. Frontend demo; backend integration is in progress. Live Demo

Architecture At A Glance

┌───────────────────────────────────────────────┐
│  entrypoints/  (CLI + HTTP API)                │  presentation
├───────────────────────────────────────────────┤
│  service/      (use cases: memorize/retrieve)  │  application
├───────────────────────────────────────────────┤
│  memory/       (extract + search + cascade)    │  domain
├───────────────────────────────────────────────┤
│  infra/        (markdown / sqlite / lancedb)   │  infrastructure
└───────────────────────────────────────────────┘
        ↑                    ↑
   component/            core/
   (LLM/Embedding)       (observability/lifespan)

DDD 5 layers, single-direction dependency. See docs/architecture.md.

Storage Layout

~/.everos/
├── default_app/                  # app_id  ("default" → "default_app" on disk)
│   └── default_project/          # project_id ("default" → "default_project")
│       ├── users/<user_id>/
│       │   ├── user.md           # profile
│       │   ├── episodes/         # daily-log episodes (visible)
│       │   ├── .atomic_facts/    # nested facts (dotfile-hidden)
│       │   └── .foresights/      # predictive memory (dotfile-hidden)
│       └── agents/<agent_id>/
│           ├── agent.md
│           ├── .cases/           # one task case per entry
│           └── skills/           # named procedural memories
├── .index/                       # derived indexes (rebuildable from md)
│   ├── sqlite/system.db          # state + queue + audit
│   └── lancedb/*.lance/          # vector + BM25 + scalar
└── .tmp/                         # transient working files

Open any <app>/<project>/users/<user_id>/ folder in Obsidian — your agent’s brain is just files. The dotfile directories (.atomic_facts/, .foresights/, .cases/) stay hidden by default so the visible folder is the user-facing memory surface, while extracted derivatives sit quietly alongside.

Features

• Hybrid retrieval: BM25 + cosine vector ANN + scalar filters, backed by LanceDB
• Cascade index sync: edit a .md → file watcher → entry-level diff → LanceDB sync, sub-second
• Multi-source extraction: conversations / agent trajectories / file knowledge
• Dual-track memory: user-track (Episodes / Profiles) + agent-track (Cases / Skills)
• Async-first: full asyncio, single event loop
• Multi-modal: text + small image / audio inline; large media via S3/OSS reference

Project Structure

everos/                        # repo root
├── src/everos/                # main package (src layout)
│   ├── entrypoints/           # cli + api
│   ├── service/               # use case orchestration
│   ├── memory/                # domain: extract + search + cascade + prompt_slots
│   ├── infra/                 # storage: markdown + lancedb + sqlite
│   ├── component/             # cross-cutting: llm / embedding / config / utils
│   ├── core/                  # runtime: observability / lifespan / context
│   └── config/                # configuration data + Settings schema
├── tests/                     # unit / integration / golden / fixtures
├── docs/                      # design docs
└── .claude/                   # team-shared rules + skills (auto-loaded by Claude Code)

Documentation

• docs/overview.md — Project overview & vision
• docs/architecture.md — DDD layered architecture & dependency rules
• docs/engineering.md — Engineering & dev-efficiency infrastructure (CI / tooling / Claude Code)
• docs/use-cases.md — Full use-case gallery and integration examples
• docs/migration-to-1.0.0.md — Legacy API and infrastructure migration notes
• CHANGELOG.md — Release notes
• CONTRIBUTING.md — How to contribute
• .claude/rules/ — Detailed coding conventions (auto-loaded by Claude Code)

Watch EverOS

EverOS 1.0.0 is the first release of a larger memory-system roadmap. Watch this repository for upcoming work on deeper idle-time and offline evolution, benchmark releases, and more real-world agent integrations.

Most memory systems stop at chat history, opaque profiles, or vector recall. EverOS keeps memory local, Markdown-native, auditable, and self-evolving: raw memory stays readable, derived knowledge becomes a wiki, and Reflection turns repeated experience into more useful long-term behavior.

If EverOS is useful to your agent stack, starring the repo helps more builders discover it.

Star History

EverMind Ecosystems

EverMind is an open-source ecosystem for long-term memory, self-evolving agents, and memory evaluation.

Together, these repositories form EverMind’s research-to-runtime stack: new memory methods, reusable algorithms, benchmark evidence, and practical agent integrations.

Contributing

Contributions are welcome across the whole repository: architecture methods, benchmark coverage, use-case examples, documentation, and bug fixes. Browse Issues to find a good entry point, then open a PR when you are ready.

Welcome all kinds of contributions 🎉

Help make EverOS better. Code, documentation, benchmark reports, use-case write-ups, and integration examples are all valuable. Share your projects on social media to inspire others.

Connect with one of the EverOS maintainers @elliotchen200 on 𝕏 or @cyfyifanchen on GitHub for project updates, discussions, and collaboration opportunities.

Code Contributors

License

Apache License 2.0 — see NOTICE for third-party attributions.

Citation

If you use EverOS in research, see CITATION.md.