snapshot-mcp-server
Save conversation state, resume work across sessions, and organize multiple snapshots with minimal token overhead.
README
Snapshot MCP Server
A Model Context Protocol (MCP) server for saving and resuming Claude conversations with token-efficient context preservation.
What is this?
A modular snapshot tool that works standalone or integrates with larger memory systems. Save conversation state, resume work across sessions, and organize multiple snapshots with minimal token overhead.
Key features:
- Token-efficient continuation prompts (~20-30% reduction vs raw context)
- Structured or simple context input
- Works with Claude Desktop, Claude Code, Cursor, any MCP client
- Designed for composability with memory/personality systems
Installation
Quick Start (Recommended)
Step 1: Run the installer:
npx @whenmoon-afk/snapshot-mcp-server
Step 2: Restart Claude Desktop (quit and reopen completely)
That's it! The installer automatically configures your Claude Desktop.
Manual Setup
If you prefer to configure manually, add to your Claude Desktop config file:
Config File Location:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json - Linux:
~/.config/Claude/claude_desktop_config.json
Configuration:
<details> <summary><strong>macOS/Linux</strong></summary>
{
"mcpServers": {
"snapshot": {
"command": "npx",
"args": ["-y", "@whenmoon-afk/snapshot-mcp-server"]
}
}
}
</details>
<details> <summary><strong>Windows</strong></summary>
{
"mcpServers": {
"snapshot": {
"command": "cmd",
"args": ["/c", "npx", "-y", "@whenmoon-afk/snapshot-mcp-server"]
}
}
}
Note: Windows requires the cmd /c wrapper to execute npx properly.
</details>
Then restart Claude Desktop.
Usage
Save snapshot:
Save a snapshot with:
- summary: "Built REST API"
- context: "Express server, MongoDB, all CRUD working"
- next_steps: "Add auth"
Resume work:
Load latest snapshot
Named snapshots:
Save snapshot named "v1-complete" with summary: "MVP deployed" and context: "..."
Load snapshot named "v1-complete"
Manage:
List all snapshots
Delete snapshot 5
Features
- Token-efficient prompts - Pre-generated, optimized continuation prompts
- Flexible context - Simple strings or structured objects
- Named snapshots - Save milestones for easy retrieval
- Cross-client - Works with Claude Desktop, Claude Code, Cursor, etc.
- Composable - Integrates with memory/personality systems
- Local storage - Your data stays on your machine (SQLite)
Structured context example:
context: {
files: ["src/auth.ts"],
decisions: ["JWT with 24h expiration"],
blockers: ["Need email verification"],
code_state: { tests_passing: true }
}
Tools
save_snapshot- Save conversation stateload_snapshot- Resume from snapshot (latest, by ID, or by name)list_snapshots- View all snapshotsdelete_snapshot- Remove snapshots
Example Workflow
End of day:
Save a snapshot:
- summary: "Implemented JWT auth"
- context: "Login/signup working, JWT middleware done, tests passing"
- next_steps: "Add password reset"
Next session:
Load latest snapshot
Claude resumes with full context.
Troubleshooting
Tools not showing?
- Restart Claude Desktop completely (quit and reopen)
- Verify config file is correct
- Ensure Node.js 18+ is installed:
node --version
Server issues?
- Reinstall:
npx @whenmoon-afk/snapshot-mcp-server - Check client logs for errors
- Reset: delete config entry and reinstall
Technical Details
Stack: SQLite + TypeScript + MCP SDK 1.0.4 (Node.js 18+)
Database location:
- macOS:
~/.claude-snapshots/snapshots.db - Windows:
%APPDATA%/claude-snapshots/snapshots.db - Linux:
~/.local/share/claude-snapshots/snapshots.db
Development:
git clone https://github.com/WhenMoon-afk/snapshot-mcp-server.git
cd snapshot-mcp-server
npm install
npm run build
License
MIT
Contributing
Issues and PRs welcome! See CHANGELOG.md for version history.
Modular conversation state for AI assistants.
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。