Taskflow MCP
Automates Notion task workflows with AI-powered tag processing, enabling Claude to interrogate, expand, critique, rewrite, and implement tasks through intelligent tag-based commands.
README
taskflow-mcp
AI-powered Notion task workflow automation for Claude Code
taskflow-mcp is a Model Context Protocol (MCP) server that automates Notion task workflows with AI-powered tag processing. Simply add MCP tags to your Notion tasks, and Claude Code will intelligently interrogate, expand, critique, rewrite, and even implement them.
🎯 Features
Full MCP Implementation
- 4 Tools - Direct task processing, querying, and updates
- 3 Resources - Quick access to Ready, In Progress, and all tagged tasks
- 6 Prompts - AI-powered workflows (interrogate, expand, critique, user stories, rewrite, prepare for coding)
- Sampling - LLM-powered analysis integrated via prompts
Intelligent Tag Processing
- Priority-based - Tags processed in optimal order automatically
- Persona support - "think like X" tags add expert context to all analysis
- Automatic tag removal - Tags cleared after successful processing
- Dry-run mode - Preview what would happen without making changes
AI-Powered Workflows (with Automatic Actions)
- interrogate - Claude asks clarifying questions → Saves as comment
- expand - Adds technical detail, APIs, edge cases → Saves as comment
- rewrite - Rewrites task for clarity → Updates task description
- estimate - Estimates effort and refactoring → Saves as comment
- critique - Provides constructive feedback → Saves as comment
- user stories - Generates user stories → Appends to description
- to-do - Adds to todo list → No code written
- code - Actually implements the feature (WRITES CODE)
- confirm - Verifies implementation complete → Saves verification as comment
🚀 Quick Start
1. Install
git clone https://github.com/themightyboosh/taskflow-mcp.git
cd taskflow-mcp
pnpm install
2. Configure
Create .env file in your project directory:
NOTION_TOKEN=ntn_xxxxxxxxxxxxxxxxxxxxxxxxxxxxx
NOTION_DATABASE_ID=xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Get your Notion token: https://www.notion.so/my-integrations
3. Setup Claude Code
Add to your .mcp.json (or Claude Code config):
{
"mcpServers": {
"taskflow-mcp": {
"type": "stdio",
"command": "node",
"args": ["/absolute/path/to/taskflow-mcp/packages/mcp/index.js"]
}
}
}
4. Add MCP Tags in Notion
Add a multi-select property called "MCP" to your Notion database, then tag tasks with:
- think like [persona] - e.g., "think like a security engineer"
- interrogate - Claude asks clarifying questions
- expand - Add technical detail
- rewrite - Rewrite for clarity
- critique - Get constructive feedback
- user stories - Generate user stories
- to-do - Add to todo list only
- code - Implement the task automatically
5. Use in Claude Code
Via Resources:
Show me notion://tasks/ready
Via Tools:
Process all my Notion tasks
Via Prompts:
Interrogate task [task-id]
📖 Documentation
- SETUP.md - Detailed setup guide
- MCP_CAPABILITIES.md - Full MCP capability reference
🏗️ Architecture
taskflow-mcp/
├── packages/mcp/
│ ├── index.js # Main MCP server
│ ├── notion-client.js # Notion API wrapper
│ ├── utils/
│ │ ├── config-loader.js # Per-project .env loading
│ │ ├── tag-manager.js # Tag priority & processing
│ │ └── image-downloader.js # Image download for vision
├── .env.example
├── .mcp.json.example
└── README.md
🎨 Tag Processing Order
Tags are processed in this priority order:
- think like [X] - Sets persona context for all subsequent processing
- interrogate - Asks clarifying questions
- rewrite - Rewrites description for clarity
- estimate - Estimates level of effort and refactoring needs
- expand - Adds technical detail
- critique - Provides feedback (with persona if set)
- user stories - Generates user stories (appended to task description)
- to-do - Adds to todo list (no implementation)
- code - Triggers implementation (ONLY tag that writes code)
- confirm - Verifies implementation is complete
Each tag is removed after successful processing.
💡 Usage Examples
Example 1: Requirements Gathering
Notion Task: "Add user authentication"
MCP Tags: interrogate, expand, user stories
What happens:
- interrogate → Claude asks questions about auth method, session management → Saves as comment
- expand → Adds technical detail (JWT vs session, password requirements) → Saves as comment
- user stories → Generates user stories for different auth scenarios → Appends to task description
- All three tags removed after processing
- All actions performed automatically - no manual intervention needed
Example 2: Expert Review
Notion Task: "Implement payment processing"
MCP Tags: think like a security engineer, critique, expand
What happens:
- think like → Sets persona: "security engineer"
- critique → Critiques from security perspective (PCI compliance, data protection) → Saves as comment
- expand → Expands with security-focused technical details → Saves as comment
- All tags removed
- All comments automatically added to Notion task
Example 3: Full Implementation
Notion Task: "Add dark mode toggle"
MCP Tags: interrogate, expand, code
What happens:
- interrogate → Claude asks about implementation approach, storage → Saves as comment
- expand → Adds technical detail (CSS variables, localStorage, toggle component) → Saves as comment
- code → Reviews requirements → WRITES THE ACTUAL CODE (creates files, edits components, etc.)
- All tags removed after processing
- Feature is fully implemented - not just analyzed
🔧 MCP Capabilities
Tools (4)
| Tool | Description |
|---|---|
process_tasks |
Process all tasks with MCP tags |
query_tasks |
Query tasks by status or MCP tags |
add_comment |
Add comment to Notion task |
update_task |
Update task properties |
Resources (3)
| Resource | Description |
|---|---|
notion://tasks/ready |
Ready tasks with MCP tags |
notion://tasks/in-progress |
In Progress tasks with MCP tags |
notion://tasks/with-mcp-tags |
All tasks with MCP tags |
Prompts (6)
| Prompt | Description |
|---|---|
interrogate_task |
Ask clarifying questions |
expand_task |
Add technical detail |
critique_task |
Provide constructive feedback |
generate_user_stories |
Generate user stories |
rewrite_task |
Rewrite for clarity |
prepare_for_coding |
Analyze and implement |
See MCP_CAPABILITIES.md for full details.
🛠️ Development
# Install dependencies
pnpm install
# Start MCP server (for testing)
pnpm start
# Development mode (auto-reload)
pnpm dev
🐛 Troubleshooting
"No .env file found"
- taskflow-mcp requires per-project
.envfile - Run from directory containing
.envwithNOTION_TOKENandNOTION_DATABASE_ID
"Property MCP does not exist"
- Add a multi-select property called "MCP" to your Notion database
- Configure your Notion integration to have access to the database
MCP server not connecting
- Verify absolute path in
.mcp.jsonis correct - Ensure Node.js 20+ is installed
- Check Notion token is valid
- Restart Claude Code after config changes
📝 Requirements
- Node.js 20+
- pnpm 10+ (or npm)
- Notion integration with database access
- Claude Code with MCP support
🤝 Contributing
Contributions welcome! Please open an issue or PR.
Development Setup
- Fork and clone
pnpm install- Create a branch:
git checkout -b feat/my-feature - Make changes and test
- Commit:
git commit -m "feat: add my feature" - Push:
git push origin feat/my-feature - Open a PR
📧 Support
- Issues: GitHub Issues
- Discussions: GitHub Discussions
📜 License
MIT © Monumental
🎯 Roadmap
- [ ] Multiple database support
- [ ] Custom tag definitions via config
- [ ] Background polling mode (watch for new tags)
- [ ] Webhook support for real-time processing
- [ ] Analytics dashboard
- [ ] Slack/Discord notifications
- [ ] Task templates
- [ ] Batch operations
Made with ❤️ by Monumental
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。