Using Wick as an AI Agent Host

Just want to run AI agents? You don't need Go, wick init, or any framework knowledge. Download the binary (or pull the Docker image) and you're done.

Wick ships as a single self-contained binary. If your goal is to run Claude / Codex / Gemini as a Slack bot, Telegram bot, or web-based AI assistant — you only need that binary.

The same binary supports two run modes:

Mode	Command	Best for
System tray	./wick (no args)	Desktop / local machine — right-click menu, icon shows server/worker state, auto-start on login
Headless server	./wick server	Remote server, Docker, CI — no GUI, logs to stdout

Download the binary

Head to the releases page and grab the latest binary for your OS:

OS	File
Linux (amd64)	wick-linux-amd64
Linux (arm64)	wick-linux-arm64
macOS (amd64)	wick-darwin-amd64
macOS (arm64)	wick-darwin-arm64
Windows	wick-windows-amd64.exe

Option A — Desktop (system tray)

Best for running on your local machine or a team member's laptop.

# macOS / Linux
chmod +x wick
./wick setup   # first-boot: generates credentials + SQLite DB
./wick         # launches system tray — right-click to start server/worker

# Windows — double-click wick.exe, or from terminal:
wick.exe setup
wick.exe

The tray icon shows server/worker state at a glance. Right-click → Start server to bring up the web UI at http://localhost:9425.

Tray features:

• Start / stop server and worker without a terminal
• MCP install/uninstall for Claude Desktop, Cursor, Gemini CLI, Codex CLI
• Self-updater — checks for new releases on launch, downloads in background, applies on restart
• Auto-start at login (opt-in from Preferences menu)
• Per-day log rotation, "Open logs" in About menu

Option B — Headless server

Best for remote servers, Docker, or anywhere there's no desktop.

chmod +x wick
./wick setup    # first-boot init — generates credentials, SQLite DB
./wick server   # start the web UI + agent host, logs to stdout

The web UI is at http://localhost:9425. Initial credentials are printed to the terminal and saved to ~/.wick/INITIAL_CREDENTIALS.txt.

Run with Docker

Docker always runs headless (wick server under the hood).

docker run -d \
  --name wick \
  -p 9425:9425 \
  -v wick-data:/root/.wick \
  ghcr.io/yogasw/wick:latest

Persist your data

Mount /root/.wick to a volume or bind mount. That directory holds the SQLite database, sessions, workspaces, and credentials. Without a mount, everything is lost on container restart.

For production, pass env vars to override defaults:

docker run -d \
  --name wick \
  -p 9425:9425 \
  -v wick-data:/root/.wick \
  -e APP_BASE_URL=https://wick.example.com \
  -e APP_ADMIN_EMAILS=you@example.com \
  ghcr.io/yogasw/wick:latest

Docker Compose

services:
  wick:
    image: ghcr.io/yogasw/wick:latest
    ports:
      - "9425:9425"
    volumes:
      - wick-data:/root/.wick
    environment:
      APP_BASE_URL: https://wick.example.com
      APP_ADMIN_EMAILS: you@example.com
    restart: unless-stopped

volumes:
  wick-data:

What to do next

1. Add a provider — go to /tools/agents → Providers, point wick at your Claude / Codex / Gemini binary and your PAT.
2. Connect a channel — Slack (Socket Mode bot token), Telegram bot token, or just use the always-on Web UI.
3. Create a workspace — a folder on disk the agent uses as its working directory. The default workspace is created automatically.
4. Start a conversation — send a message in the web UI, Slack thread, or Telegram chat. Wick spawns the agent and routes the conversation.

See AI Agents for the full breakdown, or jump to:

• Providers — binary path, PAT, multi-instance
• Channels — Slack, Telegram, Web UI setup
• Workspaces — workspace management
• Command Gate — what the gate sidecar does and why it matters