Slack MCP

Slack MCP

Read and respond to Slack from Claude. Works as an MCP server (for Claude Desktop/Claude Code) and as a standalone CLI (for GitHub Actions).

Features

• Multi-workspace support - Configure multiple Slack workspaces
• Read tools - Summarize activity, list channels, read messages, search
• Write tools - Send messages, reply to threads, add reactions
• Standalone CLI - Generate markdown summaries for GitHub Actions

Installation

1. Create Slack App(s)

For each workspace you want to connect:

1. Go to https://api.slack.com/apps → Create New App → From manifest
2. Select your workspace
3. Paste this manifest:

display_information:
  name: Claude Slack Reader
  description: Read and respond to Slack from Claude
oauth_config:
  scopes:
    user:
      # Read
      - channels:history
      - channels:read
      - groups:history
      - groups:read
      - im:history
      - im:read
      - mpim:history
      - mpim:read
      - users:read
      - search:read       # slack_search
      # Write
      - chat:write        # send messages to channels + DMs
      - reactions:write   # add emoji reactions
      - files:write       # upload files / images
      - im:write          # open a NEW DM with someone you've never messaged (existing DMs work without this)
settings:
  org_deploy_enabled: false
  socket_mode_enabled: false
  token_rotation_enabled: false

1. Click Install to Workspace → Allow
2. Copy the User OAuth Token (starts with xoxp-)

Note: Slack requires a separate app per workspace (public distribution requires Slack approval).

2. Configure Tokens

Create ~/.mcp-auth/slack/config.json:

{
  "workspaces": {
    "work": {
      "name": "Work Slack",
      "token": "xoxp-your-token-here",
      "priority": 1
    },
    "research": {
      "name": "Research Group",
      "token": "xoxp-your-token-here",
      "priority": 2
    }
  },
  "default_workspace": "work"
}

Or use environment variables: ``bash export SLACK_USER_TOKEN=xoxp-... # Single workspace export SLACK_TOKEN_WORK=xoxp-... # Multiple workspaces export SLACK_TOKEN_RESEARCH=xoxp-... ``

3. Install Dependencies

cd ~/src/slack-mcp
pip install -r requirements.txt

4. Add to Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "slack": {
      "command": "python3",
      "args": ["/path/to/slack-mcp/server.py"]
    }
  }
}

Restart Claude Desktop.

Project Structure

slack-mcp/
├── src/
│   ├── config.py          # Multi-workspace configuration
│   ├── slack_client.py    # Slack API wrapper (read + write)
│   ├── summarizer.py      # Message categorization
│   └── mcp_server.py      # MCP server with tools
├── server.py              # Entry point for MCP server
├── slack_summary.py       # Standalone CLI
├── requirements.txt
├── pyproject.toml
└── .github/
    └── workflows/
        └── daily-summary.yml

MCP Tools

Read Tools

| Tool | Description | |------|-------------| | slack_summary | Overview of DMs, mentions, channel activity. Use mode: "quick" (default) or mode: "full" | | slack_channels | List all channels (filter by type: all, channels, dms, groups) | | slack_channel | Read messages from a specific channel | | slack_thread | Read messages in a thread | | slack_search | Search messages | | slack_unread | Get unread message counts | | slack_workspaces | List configured workspaces |

Write Tools

| Tool | Description | |------|-------------| | slack_send | Send message to channel or DM | | slack_reply | Reply in a thread | | slack_react | Add emoji reaction |

Example Usage in Claude

"What's happening in my Slack?"
→ Uses slack_summary with quick mode

"Show me the #engineering channel"
→ Uses slack_channel

"Reply to that thread saying I'll review it tomorrow"
→ Uses slack_reply

"Add a thumbsup to that message"
→ Uses slack_react

Standalone CLI

Generate markdown summaries without Claude:

# Print summary to stdout
python slack_summary.py

# Save to file
python slack_summary.py --output slack-summary.md

# Look back 48 hours
python slack_summary.py --hours 48

# Specific workspace
python slack_summary.py --workspace work

# Action items only
python slack_summary.py --action-items-only

GitHub Actions Workflow

The included workflow runs daily and commits a summary:

# .github/workflows/daily-summary.yml
name: Daily Slack Summary

on:
  schedule:
    - cron: '0 12 * * *'  # 7 AM EST
  workflow_dispatch:

jobs:
  summarize:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-python@v5
        with:
          python-version: '3.11'
      - run: pip install -r requirements.txt
      - env:
          SLACK_USER_TOKEN: ${{ secrets.SLACK_USER_TOKEN }}
        run: python slack_summary.py --output slack-summary.md
      - run: |
          git config user.email "github-actions[bot]@users.noreply.github.com"
          git config user.name "github-actions[bot]"
          git add slack-summary.md
          git diff --quiet --staged || git commit -m "Daily Slack summary [skip ci]"
          git push

Add your token as a repository secret: Settings → Secrets → Actions → New repository secret → SLACK_USER_TOKEN

Configuration Sync

To sync config across machines, symlink to a cloud folder:

mkdir -p ~/.mcp-auth/slack
ln -sf ~/Dropbox/mcp-auth/slack/config.json ~/.mcp-auth/slack/config.json
# or Box, iCloud, etc.

Performance

• Quick mode (default): ~4 seconds - scans recent DMs and channels
• Full mode: ~20 seconds - detailed scan of all activity
• Cached calls: ~1.5 seconds - conversation list cached for 5 minutes

Requirements

• Python 3.10+
• slack-sdk - Slack API client
• mcp - Model Context Protocol server