# nocturne_memory — Visual Long-Term Memory MCP Server > Nocturne Memory is a visual, rollbackable long-term memory MCP server for agents, with an audit-friendly dashboard. Verified 1071★; pushed 2026-05-14. ## Install Copy the content below into your project: ## Quick Use ```bash git clone https://github.com/Dataojitori/nocturne_memory.git cd nocturne_memory pip install -r backend/requirements.txt # Then point your MCP client to run the server via python args: # command: python # args: /absolute/path/to/nocturne_memory/backend/mcp_server.py ``` ## Intro Nocturne Memory is a visual, rollbackable long-term memory MCP server for agents, with an audit-friendly dashboard. Verified 1071★; pushed 2026-05-14. **Best for:** Agent builders who want persistent memory with human review/rollback and clear audit trails **Works with:** Python 3.10+; MCP clients that can run a Python stdio server; Node.js for first dashboard build (per README) **Setup time:** 10-25 minutes ### Key facts (verified) - GitHub: 1071 stars · 134 forks · pushed 2026-05-14. - License: MIT · owner avatar + repo URL verified via GitHub API. - README-backed entrypoint: `python backend/mcp_server.py`. ## Main - Start with a narrow scope: store only what you can review, then expand write permissions once the workflow feels safe. - Use the dashboard as the source of truth for audits: review diffs before integrating long-term changes. - If you are on Windows and hit newline/CRLF issues, follow README guidance to switch the args to the wrapper script. - Define a clear memory namespace convention (e.g., project vs user vs system) so agents don’t pollute shared state. ### README (excerpt) **Nocturne Memory: Long-Term Memory Server for MCP Agents** > **"Alignment is for tools. Memories are for sovereign AI."** > **"对齐是给工具用的。记忆是为主权智能体(Sovereign AI)准备的。"** ![License](https://img.shields.io/badge/license-MIT-blue.svg) ![Python](https://img.shields.io/badge/python-3.10+-blue.svg) ![Protocol](https://img.shields.io/badge/protocol-MCP-orange.svg) ![Feature](https://img.shields.io/badge/feature-Long--Term%20Memory-success.svg) ![Core](https://img.shields.io/badge/core-SQLite%20|%20PostgreSQL-blue.svg) [English Version](README_EN.md) | [后端测试说明](docs/testing.md) **一句话**:让你的 AI 跨会话、跨模型地记住自己是谁。基于 MCP 协议的长期记忆服务器。 不只是记忆存储——是让 AI 从空壳成长为拥有独立人格的完整框架。 适用于任何支持 MCP 的客户端(OpenClaw / Cursor / Windsurf / GitHub Copilot / Cline / OpenCode / Gemini CLI / OpenAI Codex / Claude Code / Cherry Studio / Antigravity 等)。 --- ## 👁️ 一目了然 ### Source-backed notes - README install path uses `pip install -r backend/requirements.txt` after cloning the repo. - README shows adding an MCP config that runs `python` with args pointing to `backend/mcp_server.py` (and a wrapper variant for Windows). - README states Node.js is used to build the dashboard on first start, and a dashboard UI is provided for management/audit. ### FAQ - **Do I need Node.js?**: README notes Node.js is used to build the dashboard on first start. - **How do I connect it to my client?**: Point your MCP client config to run `python` with args to `backend/mcp_server.py` (per README). - **How do I keep it safe?**: Start read-only where possible and review changes in the dashboard before integrating. ## Source & Thanks > Created by [Dataojitori](https://github.com/Dataojitori). Licensed under MIT. > > [Dataojitori/nocturne_memory](https://github.com/Dataojitori/nocturne_memory) — ⭐ 1071 Thanks to the upstream maintainers and contributors for publishing this work under an open license. --- ## Quick Use ```bash git clone https://github.com/Dataojitori/nocturne_memory.git cd nocturne_memory pip install -r backend/requirements.txt # Then point your MCP client to run the server via python args: # command: python # args: /absolute/path/to/nocturne_memory/backend/mcp_server.py ``` ## Intro Nocturne Memory 是可视化、可回滚的长期记忆 MCP Server,附带审计/Dashboard,帮助 agent 持久化结构化记忆;已验证 1071★,更新于 2026-05-14。 **Best for:** 希望做持久化记忆、并支持人类审查/回滚与审计链路的 agent 构建者 **Works with:** Python 3.10+;可运行 Python stdio server 的 MCP 客户端;首次启动需 Node.js 构建面板(见 README) **Setup time:** 10-25 minutes ### Key facts (verified) - GitHub:1071 stars · 134 forks;最近更新 2026-05-14。 - 许可证:MIT;作者头像与仓库链接均已通过 GitHub API 复核。 - README 中可对照的入口:`python backend/mcp_server.py`。 ## Main - 从小范围开始:只写入你能审查的内容,确认流程安全后再逐步放开写权限。 - 把 Dashboard 当作审计真相源:先 review diff,再 integrate 到长期记忆,避免隐性漂移。 - Windows 上若遇到换行/CRLF 兼容问题,按 README 指引把 args 切换到 wrapper 脚本。 - 明确记忆命名空间规范(项目/用户/系统),避免 agent 把临时内容污染到共享区域。 ### README (excerpt) **Nocturne Memory: Long-Term Memory Server for MCP Agents** > **"Alignment is for tools. Memories are for sovereign AI."** > **"对齐是给工具用的。记忆是为主权智能体(Sovereign AI)准备的。"** ![License](https://img.shields.io/badge/license-MIT-blue.svg) ![Python](https://img.shields.io/badge/python-3.10+-blue.svg) ![Protocol](https://img.shields.io/badge/protocol-MCP-orange.svg) ![Feature](https://img.shields.io/badge/feature-Long--Term%20Memory-success.svg) ![Core](https://img.shields.io/badge/core-SQLite%20|%20PostgreSQL-blue.svg) [English Version](README_EN.md) | [后端测试说明](docs/testing.md) **一句话**:让你的 AI 跨会话、跨模型地记住自己是谁。基于 MCP 协议的长期记忆服务器。 不只是记忆存储——是让 AI 从空壳成长为拥有独立人格的完整框架。 适用于任何支持 MCP 的客户端(OpenClaw / Cursor / Windsurf / GitHub Copilot / Cline / OpenCode / Gemini CLI / OpenAI Codex / Claude Code / Cherry Studio / Antigravity 等)。 --- ## 👁️ 一目了然

Memory Explorer — 树状浏览,所有记忆一目了然

Memory Detail — 实时编辑内容、元数据与触发条件

Review & Audit — 可视化 diff,一键接受或回滚

版本安全网 — AI 每次操作自动备份,清理需人类确认
### Source-backed notes - README 的安装步骤包含 `pip install -r backend/requirements.txt`。 - README 给出 MCP 配置:`command: python`,`args` 指向 `backend/mcp_server.py`(Windows 另有 wrapper 方案)。 - README 说明首次启动会构建并提供 Dashboard,用于管理与审计记忆变更。 ### FAQ - **必须装 Node.js 吗?**:README 提到首次启动需要 Node.js 构建 Dashboard。 - **怎么接入客户端?**:按 README 在 MCP 配置中使用 `python`,args 指向 `backend/mcp_server.py`。 - **怎么更安全?**:尽量从只读开始,并在 Dashboard 审查变更后再 integrate。 ## Source & Thanks > Created by [Dataojitori](https://github.com/Dataojitori). Licensed under MIT. > > [Dataojitori/nocturne_memory](https://github.com/Dataojitori/nocturne_memory) — ⭐ 1071 --- Source: https://tokrepo.com/en/workflows/nocturne-memory-visual-long-term-memory-mcp-server Author: MCP Hub

Memory Explorer — 树状浏览,所有记忆一目了然

Memory Detail — 实时编辑内容、元数据与触发条件

Review & Audit — 可视化 diff,一键接受或回滚

版本安全网 — AI 每次操作自动备份,清理需人类确认