Project Brain
A local MCP memory server that gives AI assistants durable project memory across coding sessions, storing context, changes, and decisions.
README
Project Brain
Local MCP memory server for AI-assisted development.
What It Is
- Project Brain runs as a local MCP server via
project-brain stdio. - It gives your AI assistant durable project memory across coding sessions.
- It stores project context, changes, decisions, and progress under
.project-brain/. - The assistant should access memory through MCP tools, not by editing
.project-brain/files directly.
In Practice
| Feature | Experience |
|---|---|
| Change History | The AI becomes more "stateful" and stops repeating the same architectural mistakes. |
| Architecture Index | The AI is less likely to lose track of things during cross-file changes. |
| Preference Capture | The AI learns your architecture decisions and implementation choices, so its outputs feel more consistent¹. |
| Failed Attempt Memory | The AI stops recommending approaches you already tried and ruled out. |
Quick Start
Copy the text below into your AI assistant and let it complete the installation:
Please install Project Brain in this repository by following https://github.com/myczh-1/project-brain/docs/install.md.
Configure Project Brain as a local MCP server through `project-brain stdio`.
Do not edit `.project-brain/` files directly.
Explain which files you plan to modify before editing them.
Or, if you prefer, you can follow the docs and install it manually.
Integration Guides
Learn More
How It Works After Install
Once installed, your AI assistant will:
- connect to the local MCP server exposed by
project-brain stdio, - follow repository-local instructions from
AGENTS.mdwhen present, - read Project Brain when it needs project memory or task context,
- and update Project Brain through MCP tools instead of editing
.project-brain/files directly.
For the full workflow, see docs/guide-openspec-integration.md.
Architecture
Project Brain follows a layered architecture:
- protocol: Pure type definitions and schemas.
- core: Domain logic, commands, queries, and ports.
- infra-fs: Filesystem implementation of storage and git ports.
- mode-embedded: File-based integration helpers for repository-local workflows.
- app: Bootstrap CLI and stdio entry point.
License
MIT
<small> 1: Or the model's own preferences </small>
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。