AI-powered Slack intelligence for macOS.
Syncs your workspace locally, generates briefings, tracks action items, and analyzes team dynamics.

What is Watchtower?

Watchtower is a native macOS app that turns your Slack workspace into an actionable knowledge base. A background daemon syncs messages into a local SQLite database, then AI pipelines distill them into briefings, digests, tracks, and people analytics — all without leaving your desktop.

[Slack API] → [Local SQLite] → [AI Pipelines] → [Desktop App]
                                      ↓
                              Briefings · Digests
                              Tracks · People · Chains

Key principles: all data stays on your machine, read-only Slack access, AI runs via Claude CLI.

Features

• Daily Briefings — personalized morning overview: what needs attention, your tasks for the day, what happened, team pulse, coaching tips
• AI Chat — ask questions about your workspace in natural language, with multi-turn conversations and model selection
• Tracks — action items extracted from conversations: tasks, reviews, approvals, follow-ups with priority, status, and ownership
• Digests — channel summaries, daily rollups, weekly trends with running context that preserves topic continuity
• Chains — cross-channel discussion threads automatically linked by AI
• People Analytics — communication styles, decision roles, activity patterns, team health metrics
• Full-text Search — FTS5 search across all synced messages
• Self-improving AI — feedback loop with prompt tuning based on your ratings
• Native Notifications — alerts for new briefings, tracks, and digests
• MCP server: watchtower mcp exposes your data to any MCP client (read-only). See docs/mcp-server.md.

Install

One-liner (recommended)

curl -fsSL https://raw.githubusercontent.com/aiwathctowers/watchtower/main/scripts/install.sh | bash

Installs the desktop app to /Applications and the watchtower CLI to your PATH.

From source

Requires Go 1.25+, Swift 5.10+, macOS 14+.

git clone https://github.com/aiwathctowers/watchtower.git
cd watchtower
make app          # Full release build → build/Watchtower.app
# or
make app-dev      # Fast dev build

Pre-built binaries

Download from Releases (macOS Apple Silicon).

Getting Started

# 1. Login via Slack OAuth (opens browser)
watchtower auth login

# 2. Start the background daemon
watchtower sync --daemon

# 3. Open Watchtower.app — data appears automatically

Prerequisites:

• Slack — OAuth login handled automatically
• Claude CLI — install Claude Code for AI features (or set ANTHROPIC_API_KEY)

How It Works

The daemon (watchtower sync --daemon) polls Slack and runs five AI pipelines in sequence after each sync:

1. Digests — channel summaries with running context
2. Tracks — personal action items for the current user
3. Chains — cross-channel discussion linking
4. People — team member profiles from interaction patterns
5. Briefings — daily aggregation of all above (once per day)

The desktop app reads the same SQLite database via GRDB and updates in real-time.

Configuration

Config file: ~/.config/watchtower/config.yaml

sync:
  poll_interval: "15m"
  workers: 5
  initial_history_days: 30
digest:
  enabled: true
  model: "claude-haiku-4-5-20251001"
briefing:
  enabled: true
  hour: 8

Settings are also editable from the desktop app (Settings tab).

Data Storage

What	Where
Database	~/.local/share/watchtower/<workspace>/watchtower.db
Config	~/.config/watchtower/config.yaml
Logs	~/.local/share/watchtower/<workspace>/watchtower.log

All data is local. SQLite with WAL mode for concurrent access. The desktop app and daemon share the same database.

CLI Reference

The CLI provides full access to all features and is required for the daemon:

watchtower sync [--daemon|--full]   # Sync Slack data
watchtower ask "<question>"         # AI query
watchtower digest                   # View digests
watchtower tracks                   # View action items
watchtower briefing                 # View daily briefing
watchtower people [@user]           # People analytics
watchtower chains                   # Discussion chains
watchtower config set <key> <val>   # Configure
watchtower feedback <good|bad> ...  # Rate AI output
watchtower tune [--apply]           # Improve prompts via AI

Development

make build        # Build Go CLI only
make test         # Go tests
make test-swift   # Swift tests (395 tests)
make lint-all     # Go + Swift linting
make app-dev      # Fast dev build (CLI + desktop)
make app          # Release build with notarization

License

MIT — see LICENSE.