Slack Task MCP
An MCP server that helps users with ADHD traits manage Slack mentions by analyzing requests, structuring replies, and breaking tasks into small, actionable steps.
README
<p align="center"> <h1 align="center">Slack Task MCP</h1> </p>
<p align="center"> <strong>Zero friction from mention to action</strong> </p>
<p align="center"> <a href="https://www.npmjs.com/package/@ignission/slack-task-mcp"><img src="https://img.shields.io/npm/v/@ignission/slack-task-mcp" alt="npm version"></a> <a href="https://opensource.org/licenses/ISC"><img src="https://img.shields.io/badge/License-ISC-blue.svg" alt="License: ISC"></a> </p>
<p align="center"> An MCP server designed for users with ADHD traits </p>
<p align="center"> <a href="README.ja.md">日本語</a> </p>
Why
When Slack mentions pile up, it's hard to know where to start. Complex requests leave you frozen, unsure what to ask. Writing replies takes forever. Sound familiar?
| Problem | Solution |
|---|---|
| "What do they want?" is unclear | Clarify the purpose |
| "What should I ask?" is unknown | Identify unknowns + draft confirmation messages |
| "How should I reply?" takes time | Edit & structure reply messages |
| "What's next?" causes paralysis | Provide next actions |
Before / After
❌ Before
Mention arrives → "What do they want?" → Freeze → Postpone → Forget
✅ After
Mention arrives → Paste thread URL → Purpose, unknowns & next action organized → Start immediately
Who is this for?
- 💭 You read a Slack message multiple times but still don't know where to start
- 📝 Writing replies takes forever because you overthink the wording
- 🔄 You postpone complex requests and end up forgetting them
- 🧠 You have ADHD traits and struggle with task initiation
Use Cases
1. Break down a vague request
"Can you handle the quarterly report?"
What this tool does:
- Clarifies what "handle" means (create? review? present?)
- Identifies unknowns (deadline? format? audience?)
- Generates a confirmation message to ask the right questions
- Breaks the task into 5-minute steps once clarified
2. Structure a messy reply
"Done with the thing. Attached. Check it."
After editing:
"I've completed the quarterly report and attached it. Could you review it when you have a moment? Let me know if any changes are needed."
- Conclusion → Reasoning → Action structure
- Polite but concise tone
3. Resume work after interruption
"What was I working on for that client request?"
What this tool does:
- Searches past Slack threads
- Shows task progress and next steps
- No need to re-read the entire thread
4. Triage multiple mentions
When you have 10+ unread mentions:
- Analyze each thread quickly
- Identify priorities (blocking others? deadline?)
- Generate confirmation messages for unclear ones
- Save as tasks and tackle one by one
Demo
You: Analyze this thread https://xxx.slack.com/archives/C123/p456
Claude: ## Request Analysis
### Summary
- **Purpose**: Create sales report for Q4
- **Deliverable**: PDF report with charts
- **Deadline**: Next Friday
### Unknowns
- ❓ Which metrics to include?
- Impact: Can't start without this
- Options: Revenue only / Revenue + costs / Full P&L
### Suggested Confirmation Message
"Thanks for the request! Quick question - should I include just revenue,
or the full P&L breakdown? Also, should I use the same template as Q3?"
### Next Action
📌 Send confirmation message (2 min)
---
You: They said full P&L, same template. Break it down and save.
Claude: ✅ Task saved!
### Q4 Sales Report
1. ☐ Open Q3 template and duplicate (2 min)
2. ☐ Export Q4 data from dashboard (3 min)
3. ☐ Paste revenue numbers (5 min)
4. ☐ Paste cost numbers (5 min)
5. ☐ Update charts (5 min)
6. ☐ Write summary paragraph (5 min)
7. ☐ Export as PDF and send (2 min)
📌 Start with step 1 - just 2 minutes!
Architecture
┌─────────────────────────────────────────────────────────────┐
│ Claude Desktop / Claude Code │
│ └── MCP Client │
└──────────────────────────┬──────────────────────────────────┘
│ MCP Protocol
┌──────────────────────────▼──────────────────────────────────┐
│ Slack Task MCP Server │
│ ┌────────────────────────────────────────────────────────┐ │
│ │ MCP Tools │ │
│ │ • get_slack_thread • save_task • analyze_request │ │
│ │ • search_slack • list_tasks • draft_reply │ │
│ │ • complete_step │ │
│ └────────────────────────────────────────────────────────┘ │
│ ┌────────────────────────────────────────────────────────┐ │
│ │ Agent SDK Layer (AI-powered analysis) │ │
│ │ • analyze.js → Request analysis │ │
│ │ • draft-reply.js → Reply editing │ │
│ └────────────────────────────────────────────────────────┘ │
└──────────────────────────┬──────────────────────────────────┘
│
┌────────────────┴────────────────┐
▼ ▼
┌─────────────┐ ┌─────────────┐
│ Slack API │ │ Claude API │
│ (User Token)│ │ (Agent SDK) │
└─────────────┘ └─────────────┘
Features
| Tool | Description | Agent SDK |
|---|---|---|
get_slack_thread |
Fetch messages from a Slack thread URL | - |
analyze_request |
Analyze requests, identify purpose & unknowns | ✅ |
draft_reply |
Edit and structure replies logically | ✅ |
save_task |
Save tasks (broken into <5 min steps) | - |
list_tasks |
Show active task list | - |
search_tasks |
Search tasks by keyword/date | - |
complete_step |
Mark a step as complete | - |
search_slack |
Search Slack messages by keyword | - |
Quick Start
1. Authenticate with Slack
npx -y @ignission/slack-task-mcp auth login
A browser will open for Slack authentication. Repeat for multiple workspaces.
2. Configure Claude Code / Claude Desktop
Claude Code (Terminal):
claude mcp add slack-task -- npx -y @ignission/slack-task-mcp
Claude Desktop (~/.claude/claude_desktop_config.json):
{
"mcpServers": {
"slack-task": {
"command": "npx",
"args": ["-y", "@ignission/slack-task-mcp"]
}
}
}
3. Restart
Restart Claude Code / Claude Desktop to apply the configuration.
Usage
Basic Workflow
1. get_slack_thread → Fetch thread (context DB)
2. analyze_request → Generate purpose, unknowns, confirmation drafts
3. draft_reply → Edit and structure reply drafts
4. save_task → Save as task
5. complete_step → Track progress
Examples
Fetch & Analyze a Thread
Analyze this Slack thread:
https://xxx.slack.com/archives/C12345678/p1234567890123456
Save a Task
Break this down into steps under 5 minutes and save it
Edit a Reply
Edit this reply: "Report is done. Attached. Please check."
List Tasks
Show my task list
Complete a Step
Mark step 1 as complete
ADHD-Friendly Design
- Offload from your brain — Save tasks and forget them. Free up working memory.
- Break into <5 min steps — Build small wins
- Easiest step first — Lower the barrier to start
- Clear stopping points — Easy to pause and resume
- Slack as context DB — Ask Claude "what happened with that?"
The Real Value: Stop Carrying Everything in Your Head
❌ Keeping tasks in your head
"Don't forget this" + "That too" + "And this..." → Constant working memory drain → Exhaustion
✅ Offload to MCP
Save task → OK to forget → Brain freed up → Focus on just one thing
Your brain is for thinking, not for storing. Let the tool remember so you don't have to.
Tech Stack
| Technology | Purpose |
|---|---|
| Node.js (ES Modules) | Runtime |
| MCP Protocol | Communication with Claude Code/Desktop |
| Claude Agent SDK | AI-powered request analysis & reply editing |
| Slack Web API | Slack integration (User Token) |
| Zod | Schema validation |
| Cloudflare Workers | OAuth authentication (token exchange) |
Project Structure
slack-task-mcp/
├── src/
│ ├── index.js # MCP server entry point
│ ├── cli.js # CLI commands
│ ├── auth.js # OAuth authentication (hybrid)
│ ├── credentials.js # Credential management
│ ├── paths.js # Path management (XDG compliant)
│ └── agents/ # Agent SDK agents
│ ├── index.js # Common settings
│ ├── analyze.js # Request analysis
│ └── draft-reply.js # Reply editing
├── worker/ # Cloudflare Workers (token exchange)
│ ├── index.js
│ └── wrangler.toml
└── package.json
Data Storage
Data is stored following XDG Base Directory Specification:
~/.local/share/slack-task-mcp/
├── credentials/
│ ├── T01234567.json # Per-workspace credentials
│ └── T98765432.json
└── tasks.json # Task data
If XDG_DATA_HOME is set, that path will be used instead.
Troubleshooting
Slack API Errors
npx -y @ignission/slack-task-mcp auth status # Check auth status
npx -y @ignission/slack-task-mcp auth login # Authenticate new workspace
npx -y @ignission/slack-task-mcp auth logout # Logout from all workspaces
npx -y @ignission/slack-task-mcp auth logout -w mycompany # Logout from specific workspace
MCP Server Not Recognized
- Verify the config file path is correct
- Restart Claude Code / Claude Desktop
Can't Read Private Channels
- You can only read channels you're a member of
Contributing
Issues and PRs are welcome!
License
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。