MCP Servers

A collection of Model Context Protocol servers, templates, tools and more.

M
MCP Slack Briefing
by @sodoityu
GitHub Repo(1 stars)

mcp-slack-briefing sample

Created 3/16/2026
Updated about 19 hours ago
README
View Source
Repository documentation and setup instructions

Slack Daily Briefing Agent

An agentic daily briefing system that monitors Slack channels, generates AI-powered summaries using a local LLM, and answers follow-up questions — all without sending any data to external AI services.

Fork of sodoityu/mcp-slack-briefing, transformed into a fully automated, privacy-safe agentic system.

What it does

  1. Collects messages from your Slack channels via MCP (Model Context Protocol)
  2. Filters for important messages (incidents, tickets, keywords, emojis)
  3. Summarizes using a local Ollama LLM — no data leaves your machine
  4. Posts the briefing to a Slack channel (header + threaded summary)
  5. Answers follow-up questions in the briefing thread using the local LLM

Quick Start

Prerequisites

Choose your platform:

macOS (Apple Silicon M1/M2/M3, 16GB RAM)

brew install python podman ollama

Linux (Fedora / RHEL)

sudo dnf install python3 python3-pip podman
curl -fsSL https://ollama.com/install.sh | sh

1. Clone and setup

git clone https://github.com/archith-kadannapalli/mcp-slack-briefing.git
cd mcp-slack-briefing
chmod +x setup.sh && ./setup.sh

The setup script auto-detects your platform (macOS or Linux) and:

  • Creates a Python virtual environment
  • Installs Python dependencies
  • Pulls the llama3.1:8b AI model (~4.7GB, one-time)
  • Sets up Podman (initializes VM on macOS, native on Linux)
  • Pulls the Slack MCP container
  • Creates .env and .mcp.json config templates

2. Get your Slack tokens

Open Slack in your browser (not the desktop app), then open DevTools (F12):

| Token | Where to find it | |-------|-----------------| | xoxc-* | Network tab > click any channel > find request to api.slack.com > look for token in request body | | xoxd-* | Application tab > Cookies > cookie named d |

3. Configure

Edit .mcp.json — add your Slack tokens:

{
  "mcpServers": {
    "slack": {
      "env": {
        "SLACK_XOXC_TOKEN": "xoxc-YOUR-ACTUAL-TOKEN",
        "SLACK_XOXD_TOKEN": "xoxd-YOUR-ACTUAL-TOKEN",
        "SLACK_WORKSPACE_URL": "https://your-workspace.slack.com"
      }
    }
  }
}

Edit .env — add your channel IDs:

# Right-click channel in Slack > "View channel details" > ID at bottom
BRIEFING_CHANNEL_ID=C04XXXXXXXX
MONITORED_CHANNELS='[{"id":"C04XXXXXXXX","name":"your-channel"},{"id":"C01XXXXXXXX","name":"another-channel"}]'

4. Start Podman (platform-specific)

macOS:

podman machine init    # first time only
podman machine start

Linux (Fedora/RHEL):

# Podman runs natively, no setup needed
# Verify with:
podman info

5. Start Ollama

macOS:

ollama serve &
ollama pull llama3.1:8b    # first time only

Linux:

sudo systemctl start ollama
sudo systemctl enable ollama    # auto-start on boot
ollama pull llama3.1:8b         # first time only

6. Test

source venv/bin/activate
set -a; source .env; set +a

# Test collection (fetches messages from your channels)
python daily_briefing.py 24 briefing_test.txt false

# Test full pipeline (collect + summarize + post to Slack)
./run_daily_briefing.sh

7. Start Q&A listener

After a briefing is posted, start the listener to answer follow-up questions:

./start_listener.sh          # foreground (Ctrl+C to stop)
./start_listener.sh --bg     # background (logs to logs/qa_listener.log)
./stop_listener.sh           # stop background listener

Reply in the briefing thread in Slack with any question — answered within 10 seconds.

8. Automate (optional)

./setup_cron.sh

This sets up:

  • A cron job to run the daily briefing at your chosen time
  • A persistent service for the Q&A listener:
    • macOS: launchd (auto-starts on login)
    • Linux: systemd user service (auto-starts on boot, survives logout)

Architecture

Cron (daily)                    Your Machine
     |                    (everything runs locally)
     v
daily_briefing.py -----> Slack MCP (Podman) -----> Slack API
     |                                              |
     v                                              v
briefing_YYYY-MM-DD.txt              Monitored Channels
     |
     v
ollama_summarizer.py --> Ollama (localhost:11434)
     |
     v
briefing_summary_YYYY-MM-DD.txt
     |
     v
post_summary_to_slack.py ---------> Slack Channel
     |                               |
     v                               v
qa_listener.py (polls every 10s)   User replies in thread
     |                               |
     v                               v
Ollama (generates answer) -------> Reply posted in thread

Privacy and Security

No Slack data ever leaves your machine. All AI processing happens locally via Ollama.

| Safeguard | What it does | |-----------|-------------| | S1 - Channel allowlist | Only monitors channels you explicitly list in .env | | S2 - Context scoping | Q&A answers only reference the current briefing, not all of Slack | | S3/S4 - Access control | Channel search restricted to the allowlist | | S5 - Thread-only replies | Bot never posts top-level messages, only in-thread | | S6 - Local LLM only | Endpoint validated as localhost — external URLs blocked | | S7 - No hardcoded tokens | All credentials in .env / .mcp.json, both gitignored | | S8 - PII sanitization | Emails, phone numbers, IPs stripped before LLM sees them |

File Structure

mcp-slack-briefing/
├── setup.sh                  # One-time setup (run first)
├── run_daily_briefing.sh     # Full pipeline: collect + summarize + post
├── start_listener.sh         # Start Q&A listener
├── stop_listener.sh          # Stop Q&A listener
├── setup_cron.sh             # Set up daily cron + persistent service
├── daily_briefing.py         # Message collection from Slack via MCP
├── ollama_summarizer.py      # Local Ollama summarization + Q&A
├── post_summary_to_slack.py  # Post briefing to Slack channel
├── qa_listener.py            # Poll briefing thread, answer questions
├── safeguards.py             # Security checks (S1-S8)
├── .env.example              # Environment config template
├── .mcp.json.example         # MCP/Slack config template
├── requirements.txt          # Python dependencies
├── CLAUDE.md                 # AI assistant instructions
├── AGENTIC_DESIGN.md         # Architecture and design document
└── README.md                 # This file

Platform Differences

| Feature | macOS | Linux (Fedora/RHEL) | |---------|-------|---------------------| | Python install | brew install python | sudo dnf install python3 python3-pip | | Podman install | brew install podman | sudo dnf install podman (often pre-installed) | | Podman startup | podman machine init && podman machine start | Not needed (runs natively) | | Ollama install | brew install ollama | curl -fsSL https://ollama.com/install.sh \| sh | | Ollama startup | ollama serve & | sudo systemctl start ollama | | Q&A service | launchd (auto-start on login) | systemd user service (survives logout) | | Container arch | May show linux/amd64 warning on ARM | Native architecture match |

Upgrading to OpenShift

When ready to move from local Ollama to a shared OpenShift deployment:

  1. Deploy Ollama/vLLM/RHOAI on your cluster
  2. Update two env vars in .env: 
    OLLAMA_BASE_URL=https://llm.apps.your-cluster.internal
OLLAMA_MODEL=your-deployed-model
  3. Update ALLOWED_LLM_HOSTS in safeguards.py to include your cluster hostname
  4. Everything else stays the same

Troubleshooting

| Problem | Fix | |---------|-----| | externally-managed-environment error | Use the venv: source venv/bin/activate | | json.decoder.JSONDecodeError on .env | Wrap MONITORED_CHANNELS value in single quotes | | json.decoder.JSONDecodeError on .mcp.json | Check for double-double quotes around tokens | | pydantic ValidationError in output | Harmless warning from MCP server — pipeline continues | | Ollama not available | macOS: ollama serve & / Linux: sudo systemctl start ollama | | Podman machine not running (macOS) | podman machine start | | No briefing thread found | Run ./run_daily_briefing.sh first, then start listener | | Q&A not detecting replies | Reply in the briefing thread, not as a new channel message | | Poetry could not find pyproject.toml | Use venv, not poetry: source venv/bin/activate | | Linux: permission denied on scripts | chmod +x *.sh | | Linux: Ollama won't start | sudo systemctl enable --now ollama |

License

MIT License — based on sodoityu/mcp-slack-briefing

Quick Setup
Installation guide for this server

Install Package (if required)

uvx mcp-slack-briefing

Cursor configuration (mcp.json)

{ "mcpServers": { "sodoityu-mcp-slack-briefing": { "command": "uvx", "args": [ "mcp-slack-briefing" ] } } }