memos
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@memosremember that I prefer using VSCode for coding"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
agent-memory-v3
基于 MemOS(Apache-2.0)二次开发的 agent 长期记忆系统 v2, 目标是补齐 MemOS 云端版独有的「按类别抽取 + 按视图分发」管线能力(不改存储引擎), 形成 100% 本地部署、可外接自有 AI API、经 MCP 接入 Codex 等编码 agent 的记忆系统, 替代前代 agent-memory-skill(SQLite + 标签精确召回)。
修改声明:本仓库是 MemOS v2.0.34(commit
41bf5c7f)的修改版,依 Apache-2.0 许可证发布; 上游 LICENSE 原样保留,全部差异登记于PATCHES.md,基线信息见UPSTREAM.md。
一期范围:六路记忆管线
写入时按类别 key 并行抽取,读出时按类别视图分发(对齐 MemOS 云版 API 形状):
路 | 类别 key | 产出契约 | 状态 |
事实 |
|
| M1 |
偏好 |
|
| M1 |
事件 |
|
| M2 |
工具记忆 |
|
| M3 |
技能 |
|
| M4 |
模态 |
| 先解析(OCR/转文本)再进各抽取路 | M5 |
明确不做:Profile 模板路、知识库路(内置文档 RAG)、Dashboard 治理面板。
Related MCP server: memento
How to start(for your agent)
前置条件:Docker(Neo4j + Qdrant + Redis + MemOS 服务)、一个 OpenAI 兼容 LLM 端点、bge-m3 embedding。
# 部署目录在设备本地 llwwds_application/docker_file/memos/(不在本仓库内)
cd ~/llwwds_application/docker_file/memos
cp .env.example .env # 填入 LLM/embedder 端点与密钥
docker compose up -d
docker compose ps # 等待全部 healthy
# 最小验证:写入并召回
curl -s localhost:8000/product/add/message -H 'Content-Type: application/json' \
-d '{"messages":[{"role":"user","content":"我喜欢用 VSCode 写代码"}],"user_id":"test"}'
curl -s localhost:8000/product/search/memory -H 'Content-Type: application/json' \
-d '{"query":"我用什么编辑器","user_id":"test"}'MCP 接入(Codex,~/.codex/config.toml):
[mcp_servers.memos]
command = "<venv-python 路径>"
args = ["<仓库路径>/src/memos/api/mcp_serve.py"]文件夹定义
路径 | 职责 |
| 上游 MemOS 源码(基线),自有管线代码叠加其上 |
| 上游官方 Dockerfile 与 compose(部署副本在设备本地 |
| 被本项目替换的上游根文档存档(README/AGENTS) |
| 与源码版本绑定的正式文档(API 契约、部署说明) |
| 上游测试 + 本项目管线测试 |
| 上游基线登记 / 对上游的就地修改清单 |
设计思路
核心架构决策与数据流见本仓库 docs/ 与知识库权威计划文档
(obsidian_file/02 项目/021 专业项目/0211 个人项目/agent_sql_memory/计划 2026-10-03 MemOS 二次开发 v2 开发计划.md)。
一句话本质:云版没有为类别记忆加存储层,加的是「写路径上的分类抽取」+「读路径上的分类分发」; 开源版存储层字段全在、响应模型已预定义,二次开发是填空,不是造轮子。
当前状态
M0 环境与基线:进行中(基线导入完成,bug 修复与闭环验证进行中)。
里程碑与验收标准以知识库权威计划文档为准。
License
Apache-2.0(继承上游 MemOS)。
This server cannot be deployed
Maintenance
Related MCP Connectors
Persistent memory for AI agents across Claude, ChatGPT and any MCP client.
Long-term memory for AI coding agents: durable project facts, recalled by every MCP client.
- memnodeOAuthdev.memnode
Persistent, inspectable memory for AI agents with lineage, correction, and a hosted MCP endpoint.
Hosted MCP memory for coding agents: persistent across sessions, editable markdown, team sharing.
Related MCP Servers
- AlicenseNot gradedqualityBmaintenanceProvides a persistent, local-first memory for coding agents over MCP, enabling automatic recall and recording of past work, failures, and decisions to reduce repetition and token usage.MIT
- AlicenseNot gradedqualityAmaintenanceProvides persistent memory for AI coding agents via MCP, enabling agents to store and semantically recall facts, events, and lessons across sessions, all running locally without cloud dependencies.Apache 2.0
- AlicenseAqualityBmaintenanceProvides self-hosted memory management for coding agents via MCP, with local vector storage and tools for adding, searching, updating, and deleting memories.81MIT
- AlicenseNot gradedqualityBmaintenanceProvides local-first persistent memory with a typed knowledge graph and bounded multi-hop retrieval via MCP, letting coding agents and local LLM systems store, search, and recall facts across sessions without hosted services or model dependencies.MIT