Each setup below uses only what ships in the binary. Start from a working install (Quickstart), then add the pieces shown. Keys always go in $HABAKKUK_HOME/.env, never in config.yaml.

A personal assistant on WhatsApp#

An agent with its own chat in your WhatsApp, with no phone number to give up and no public URL.

In WhatsApp, open Settings > Agents > Create agent, and send it "hi".
Open its chat, then Chat info > API key, and copy the key.
Run the wizard and pick WhatsApp agent:

bash

habakkuk setup
habakkuk setup service      # Linux and macOS: write a systemd or launchd file, and print how to install it

The channel it writes looks like this:

yaml

channels:
  wagent:
    type: whatsappagent
    token_env: WHATSAPP_AGENT_KEY

Only you, the agent's creator, can write to it. See WhatsApp.

A daily briefing#

A message every weekday morning with what you asked for. Set your time zone, make sure the agent can search the web, and add a schedule.

yaml

timezone: America/Toronto
search:                     # not needed when the agent's model is on an anthropic provider,
  provider: brave           # which brings its own hosted web search
  key_env: BRAVE_API_KEY

Ask in a chat ("every weekday at 7:30, send me the top three stories about renewable energy and today's weather in Toronto"), and the agent creates the schedule with schedule_create. Or add it from the terminal:

bash

habakkuk schedules add -agent habakkuk -name briefing -when "30 7 * * 1-5" \
  -prompt "Search the news and send me the top three stories about renewable energy, then today's weather in Toronto." \
  -to wagent:creator
habakkuk schedules run <id>     # try it now

habakkuk serve must be running. See Schedules and Media and web search.

A team of business agents on WhatsApp, run from Slack#

Marketing, sales and support each have their own WhatsApp agent chat. An orchestrator answers you on Slack and delegates to them.

Create three WhatsApp agents (Settings > Agents), and send each one a message.
Create a Slack app with Socket Mode (see Slack).
Configure the agents and channels:

yaml

agents:
  orchestrator:
    model: anthropic/claude-opus-5-5     # the default toolsets include delegate and board
  marketing:
    model: anthropic/claude-sonnet-5-5
    home: wa-marketing:creator
  sales:
    model: anthropic/claude-sonnet-5-5
    home: wa-sales:creator
  support:
    model: anthropic/claude-sonnet-5-5
    home: wa-support:creator
channels:
  slack:
    type: slack
    app_token_env: SLACK_APP_TOKEN
    bot_tokens_env: [SLACK_BOT_TOKEN]
    agent: orchestrator
  wa-marketing: {type: whatsappagent, token_env: WA_MARKETING_KEY, agent: marketing}
  wa-sales:     {type: whatsappagent, token_env: WA_SALES_KEY, agent: sales}
  wa-support:   {type: whatsappagent, token_env: WA_SUPPORT_KEY, agent: support}

Ask the orchestrator on Slack "what's up with the campaign?" and it delegates to marketing. The answer comes back to Slack, and a short report lands in the Marketing chat on WhatsApp, so each agent's chat keeps the history of its own work. See Multi-agent teams.

A research agent#

A second agent with a browser and web search, that you reach with /agent research or that your main agent delegates to.

yaml

search:                            # web_search for models without hosted search
  provider: tavily
  key_env: TAVILY_API_KEY
agents:
  research:
    model: deepseek/deepseek-v4-pro
    permission: workspace          # web reads run on their own; commands ask first
    toolsets: [core, memory, browser]
browser:
  mode: launch
  headless: true

bash

habakkuk run -agent research "Compare the three most used open-source CRMs. A table, then a recommendation."

On a server, install Chromium first (apt install chromium on Debian). For long jobs, put a card on the board assigned to research: the result comes back to the conversation that asked. See Browser.

A home-automation agent#

Lights, climate and media from a chat, and a turn that starts on its own when the doorbell rings.

yaml

agents:
  habakkuk:
    toolsets: [core, memory, skills, schedule, homeassistant]
homeassistant:
  url: http://homeassistant.local:8123
  token_env: HA_TOKEN                        # a long-lived access token
  allow_domains: [light, switch, climate, media_player]
  confirm: [lock, alarm_control_panel, cover]  # every call asks you first
  triggers:
    - name: doorbell
      entity_id: binary_sensor.front_doorbell
      to: "on"
      prompt: "Someone rang the doorbell at {{.Time}}. Check the porch camera and tell me who it is."
      deliver: telegram:123456789

Then ask "turn off the living room lights" in any chat. See Integrations: Home Assistant.

A meeting note-taker#

The agent joins Google Meet, Zoom or Teams calls as a named guest, reads the live captions, and sends notes with decisions and action items.

yaml

agents:
  habakkuk:
    toolsets: [core, memory, skills, schedule, meeting]
meeting:
  name: Habakkuk (notetaker)

Ask "join https://meet.google.com/abc-defg-hij and note the decisions", or "every weekday at 9:55, join my standup at ", which becomes a schedule. See Meetings.

A coding agent in a sandbox, in your editor#

An agent whose commands run in a Docker container, with workspace snapshots before every change, used from an editor that speaks the Agent Client Protocol.

yaml

agents:
  coder:
    model: anthropic/claude-opus-5-5
    workspace: ~/src/myproject
    sandbox: docker
    docker: {image: "golang:1.27", network: false, cpus: "2", memory: 2g}

bash

habakkuk acp -agent coder        # configure your editor to run this as its agent
habakkuk checkpoints -agent coder
habakkuk rollback -agent coder <id>

/undo in a chat puts back the files the last turn changed. See Files and shell and Integrations: code editors.

A watcher that only speaks when something changes#

A script schedule runs a command without calling a model, and posts only when the script prints something. It costs nothing while there is nothing to say.

bash

cat > ~/habakkuk-workspace/check-site.sh <<'EOF'
#!/bin/sh
code=$(curl -s -o /dev/null -w '%{http_code}' https://example.com)
[ "$code" = 200 ] || echo "example.com answered $code"
EOF
chmod +x ~/habakkuk-workspace/check-site.sh

habakkuk schedules add -agent habakkuk -name site-check -when "@every 5m" \
  -run ./check-site.sh -on-output deliver -timeout 30s -to telegram:123456789

The script runs in the agent's workspace (here ~/habakkuk-workspace, as habakkuk setup writes it); -workdir DIR picks another folder. With -on-output agent -prompt "...", the output starts a turn instead, so the agent can look into it. Only you can add script schedules, from the CLI. See Script schedules.