setup

name	setup
description	Run initial MolClaw setup. Use when user wants to install dependencies, authenticate WhatsApp, register their main channel, or start the background services. Triggers on "setup", "install", "configure molclaw", or first-time setup requests.

MolClaw Setup

Run all commands automatically. Only pause when user action is required (WhatsApp authentication, configuration choices).

UX Note: When asking the user questions, prefer using the AskUserQuestion tool instead of just outputting text. This integrates with Claude's built-in question/answer system for a better experience.

1. Install Dependencies

npm install

2. Install Container Runtime

First, detect the platform and check what's available:

echo "Platform: $(uname -s)"
which container && echo "Apple Container: installed" || echo "Apple Container: not installed"
which docker && docker info >/dev/null 2>&1 && echo "Docker: installed and running" || echo "Docker: not installed or not running"

If NOT on macOS (Linux, etc.)

Apple Container is macOS-only. Use Docker instead.

Tell the user:

You're on Linux, so we'll use Docker for container isolation. Let me set that up now.

Use the /convert-to-docker skill to convert the codebase to Docker, then continue to Section 3.

If on macOS

If Apple Container is already installed: Continue to Section 3.

If Apple Container is NOT installed: Ask the user:

MolClaw needs a container runtime for isolated agent execution. You have two options:

1. Apple Container (default) - macOS-native, lightweight, designed for Apple silicon
2. Docker - Cross-platform, widely used, works on macOS and Linux

Which would you prefer?

Option A: Apple Container

Tell the user:

Apple Container is required for running agents in isolated environments.

1. Download the latest .pkg from https://github.com/apple/container/releases
2. Double-click to install
3. Run container system start to start the service

Let me know when you've completed these steps.

Wait for user confirmation, then verify:

container system start
container --version

Note: MolClaw automatically starts the Apple Container system when it launches, so you don't need to start it manually after reboots.

Option B: Docker

Tell the user:

You've chosen Docker. Let me set that up now.

Use the /convert-to-docker skill to convert the codebase to Docker, then continue to Section 3.

3. Configure Claude Authentication

Ask the user:

Do you want to use your Claude subscription (Pro/Max) or an Anthropic API key?

Option 1: Claude Subscription (Recommended)

Tell the user:

Open another terminal window and run:

claude setup-token

A browser window will open for you to log in. Once authenticated, the token will be displayed in your terminal. Either:

1. Paste it here and I'll add it to .env for you, or
2. Add it to .env yourself as CLAUDE_CODE_OAUTH_TOKEN=<your-token>

If they give you the token, add it to .env. Never echo the full token in commands or output — use the Write tool to write the .env file directly, or tell the user to add it themselves:

echo "CLAUDE_CODE_OAUTH_TOKEN=<token>" > .env

Option 2: API Key

Ask if they have an existing key to copy or need to create one.

Copy existing:

grep "^ANTHROPIC_API_KEY=" /path/to/source/.env > .env

Create new:

echo 'ANTHROPIC_API_KEY=' > .env

Tell the user to add their key from https://console.anthropic.com/

Verify:

KEY=$(grep "^ANTHROPIC_API_KEY=" .env | cut -d= -f2)
[ -n "$KEY" ] && echo "API key configured: ${KEY:0:7}..." || echo "Missing"

4. Build Container Image

Build the MolClaw agent container:

./container/build.sh

This creates the molclaw-agent:latest image with Node.js, Chromium, Claude Code CLI, and agent-browser.

Verify the build succeeded by running a simple test (this auto-detects which runtime you're using):

if which docker >/dev/null 2>&1 && docker info >/dev/null 2>&1; then
  echo '{}' | docker run -i --entrypoint /bin/echo molclaw-agent:latest "Container OK" || echo "Container build failed"
else
  echo '{}' | container run -i --entrypoint /bin/echo molclaw-agent:latest "Container OK" || echo "Container build failed"
fi

5. WhatsApp Authentication

USER ACTION REQUIRED

The auth script supports two methods: QR code scanning and pairing code (phone number). Ask the user which they prefer.

The auth script writes status to store/auth-status.txt:

• already_authenticated — credentials already exist
• pairing_code:<CODE> — pairing code generated, waiting for user to enter it
• authenticated — successfully authenticated
• failed:<reason> — authentication failed

The script automatically handles error 515 (stream error after pairing) by reconnecting — this is normal and expected during pairing code auth.

Ask the user which method to use

How would you like to authenticate WhatsApp?

1. QR code in browser (Recommended) — Opens a page with the QR code to scan
2. Pairing code — Enter a numeric code on your phone, no camera needed
3. QR code in terminal — Run the auth command yourself in another terminal

Option A: QR Code in Browser (Recommended)

Clean any stale auth state and start auth in background:

rm -rf store/auth store/qr-data.txt store/auth-status.txt
npm run auth

Run this with run_in_background: true.

Poll for QR data (up to 15 seconds):

for i in $(seq 1 15); do if [ -f store/qr-data.txt ]; then echo "qr_ready"; exit 0; fi; STATUS=$(cat store/auth-status.txt 2>/dev/null || echo "waiting"); if [ "$STATUS" = "already_authenticated" ]; then echo "$STATUS"; exit 0; fi; sleep 1; done; echo "timeout"

If already_authenticated, skip to the next step.

If QR data is ready, generate the QR as SVG and inject it into the HTML template:

node -e "
const QR = require('qrcode');
const fs = require('fs');
const qrData = fs.readFileSync('store/qr-data.txt', 'utf8');
QR.toString(qrData, { type: 'svg' }, (err, svg) => {
  if (err) process.exit(1);
  const template = fs.readFileSync('.claude/skills/setup/qr-auth.html', 'utf8');
  fs.writeFileSync('store/qr-auth.html', template.replace('{{QR_SVG}}', svg));
  console.log('done');
});
"

Then open it:

open store/qr-auth.html

Tell the user:

A browser window should have opened with the QR code. It expires in about 60 seconds.

Scan it with WhatsApp: Settings → Linked Devices → Link a Device

Then poll for completion (up to 120 seconds):

for i in $(seq 1 60); do STATUS=$(cat store/auth-status.txt 2>/dev/null || echo "waiting"); if [ "$STATUS" = "authenticated" ] || [ "$STATUS" = "already_authenticated" ]; then echo "$STATUS"; exit 0; elif echo "$STATUS" | grep -q "^failed:"; then echo "$STATUS"; exit 0; fi; sleep 2; done; echo "timeout"

• If authenticated, success — clean up with rm -f store/qr-auth.html and continue.
• If failed:qr_timeout, offer to retry (re-run the auth and regenerate the HTML page).
• If failed:logged_out, delete store/auth/ and retry.

Option B: Pairing Code

Ask the user for their phone number (with country code, no + or spaces, e.g. 14155551234).

Clean any stale auth state and start:

rm -rf store/auth store/qr-data.txt store/auth-status.txt
npx tsx src/whatsapp-auth.ts --pairing-code --phone PHONE_NUMBER

Run this with run_in_background: true.

Poll for the pairing code (up to 15 seconds):

for i in $(seq 1 15); do STATUS=$(cat store/auth-status.txt 2>/dev/null || echo "waiting"); if echo "$STATUS" | grep -q "^pairing_code:"; then echo "$STATUS"; exit 0; elif [ "$STATUS" = "authenticated" ] || [ "$STATUS" = "already_authenticated" ]; then echo "$STATUS"; exit 0; elif echo "$STATUS" | grep -q "^failed:"; then echo "$STATUS"; exit 0; fi; sleep 1; done; echo "timeout"

Extract the code from the status (e.g. pairing_code:ABC12DEF → ABC12DEF) and tell the user:

Your pairing code: CODE_HERE

1. Open WhatsApp on your phone
2. Tap Settings → Linked Devices → Link a Device
3. Tap "Link with phone number instead"
4. Enter the code: CODE_HERE

Then poll for completion (up to 120 seconds):

for i in $(seq 1 60); do STATUS=$(cat store/auth-status.txt 2>/dev/null || echo "waiting"); if [ "$STATUS" = "authenticated" ] || [ "$STATUS" = "already_authenticated" ]; then echo "$STATUS"; exit 0; elif echo "$STATUS" | grep -q "^failed:"; then echo "$STATUS"; exit 0; fi; sleep 2; done; echo "timeout"

• If authenticated or already_authenticated, success — continue to next step.
• If failed:logged_out, delete store/auth/ and retry.
• If failed:515 or timeout, the 515 reconnect should handle this automatically. If it persists, the user may need to temporarily stop other WhatsApp-connected apps on the same device.

Option C: QR Code in Terminal

Tell the user to run the auth command in another terminal window:

Open another terminal and run:

cd PROJECT_PATH && npm run auth

Scan the QR code that appears, then let me know when it says "Successfully authenticated".

Replace PROJECT_PATH with the actual project path (use pwd).

Wait for the user to confirm authentication succeeded, then continue to the next step.

6. Configure Assistant Name and Main Channel

This step configures three things at once: the trigger word, the main channel type, and the main channel selection.

6a. Ask for trigger word

Ask the user:

What trigger word do you want to use? (default: Bio)

In group chats, messages starting with @TriggerWord will be sent to Claude. In your main channel (and optionally solo chats), no prefix is needed — all messages are processed.

Store their choice for use in the steps below.

6b. Explain security model and ask about main channel type

Use the AskUserQuestion tool to present this:

Important: Your "main" channel is your admin control portal.

The main channel has elevated privileges:

• Can see messages from ALL other registered groups
• Can manage and delete tasks across all groups
• Can write to global memory that all groups can read
• Has read-write access to the entire MolClaw project

Recommendation: Use your personal "Message Yourself" chat or a solo WhatsApp group as your main channel. This ensures only you have admin control.

Question: Which setup will you use for your main channel?

Options:

1. Personal chat (Message Yourself) - Recommended
2. DM with a specific phone number (e.g. your other phone)
3. Solo WhatsApp group (just me)
4. Group with other people (I understand the security implications)

If they choose option 4, ask a follow-up:

You've chosen a group with other people. This means everyone in that group will have admin privileges over MolClaw.

Are you sure you want to proceed? The other members will be able to:

• Read messages from your other registered chats
• Schedule and manage tasks
• Access any directories you've mounted

Options:

1. Yes, I understand and want to proceed
2. No, let me use a personal chat or solo group instead

6c. Register the main channel

First build, then start the app briefly to connect to WhatsApp and sync group metadata. Use the Bash tool's timeout parameter (15000ms) — do NOT use the timeout shell command (it's not available on macOS). The app will be killed when the timeout fires, which is expected.

npm run build

Then run briefly (set Bash tool timeout to 15000ms):

npm run dev

For personal chat (they chose option 1):

Personal chats are NOT synced to the database on startup — only groups are. The JID for "Message Yourself" is the bot's own number. Use the number from the WhatsApp auth step and construct the JID as {number}@s.whatsapp.net.

For DM with a specific number (they chose option 2):

Ask the user for the phone number (with country code, no + or spaces, e.g. 14155551234), then construct the JID as {number}@s.whatsapp.net.

For group (they chose option 3 or 4):

Groups are synced on startup via groupFetchAllParticipating. Query the database for recent groups:

sqlite3 store/messages.db "SELECT jid, name FROM chats WHERE jid LIKE '[e-mail]' AND jid != '__group_sync__' ORDER BY last_message_time DESC LIMIT 40"

Show only the 10 most recent group names to the user and ask them to pick one. If they say their group isn't in the list, show the next batch from the results you already have. If they tell you the group name directly, look it up:

sqlite3 store/messages.db "SELECT jid, name FROM chats WHERE name LIKE '%GROUP_NAME%' AND jid LIKE '[e-mail]'"

6d. Write the configuration

Once you have the JID, configure it. Use the assistant name from step 6a.

For personal chats (solo, no prefix needed), set requiresTrigger to false:

{
  "JID_HERE": {
    "name": "main",
    "folder": "main",
    "trigger": "@ASSISTANT_NAME",
    "added_at": "CURRENT_ISO_TIMESTAMP",
    "requiresTrigger": false
  }
}

For groups, keep requiresTrigger as true (default).

Write to the database directly by creating a temporary registration script, or write data/registered_groups.json which will be auto-migrated on first run:

mkdir -p data

Then write data/registered_groups.json with the correct JID, trigger, and timestamp.

If the user chose a name other than Bio, also update:

1. groups/global/CLAUDE.md - Change "# Bio" and "You are Bio" to the new name
2. groups/main/CLAUDE.md - Same changes at the top

Ensure the groups folder exists:

mkdir -p groups/main/logs

7. Configure External Directory Access (Mount Allowlist)

Ask the user:

Do you want the agent to be able to access any directories outside the MolClaw project?

Examples: Git repositories, project folders, documents you want Claude to work on.

Note: This is optional. Without configuration, agents can only access their own group folders.

If no, create an empty allowlist to make this explicit:

mkdir -p ~/.config/molclaw
cat > ~/.config/molclaw/mount-allowlist.json << 'EOF'
{
  "allowedRoots": [],
  "blockedPatterns": [],
  "nonMainReadOnly": true
}
EOF
echo "Mount allowlist created - no external directories allowed"

Skip to the next step.

If yes, ask follow-up questions:

7a. Collect Directory Paths

Ask the user:

Which directories do you want to allow access to?

You can specify:

• A parent folder like ~/projects (allows access to anything inside)
• Specific paths like ~/repos/my-app

List them one per line, or give me a comma-separated list.

For each directory they provide, ask:

Should [directory] be read-write (agents can modify files) or read-only?

Read-write is needed for: code changes, creating files, git commits Read-only is safer for: reference docs, config examples, templates

7b. Configure Non-Main Group Access

Ask the user:

Should non-main groups (other WhatsApp chats you add later) be restricted to read-only access even if read-write is allowed for the directory?

Recommended: Yes - this prevents other groups from modifying files even if you grant them access to a directory.

7c. Create the Allowlist

Create the allowlist file based on their answers:

mkdir -p ~/.config/molclaw

Then write the JSON file. Example for a user who wants ~/projects (read-write) and ~/docs (read-only) with non-main read-only:

cat > ~/.config/molclaw/mount-allowlist.json << 'EOF'
{
  "allowedRoots": [
    {
      "path": "~/projects",
      "allowReadWrite": true,
      "description": "Development projects"
    },
    {
      "path": "~/docs",
      "allowReadWrite": false,
      "description": "Reference documents"
    }
  ],
  "blockedPatterns": [],
  "nonMainReadOnly": true
}
EOF

Verify the file:

cat ~/.config/molclaw/mount-allowlist.json

Tell the user:

Mount allowlist configured. The following directories are now accessible:

• ~/projects (read-write)
• ~/docs (read-only)

Security notes:

• Sensitive paths (.ssh, .gnupg, .aws, credentials) are always blocked
• This config file is stored outside the project, so agents cannot modify it
• Changes require restarting the MolClaw service

To grant a group access to a directory, add it to their config in data/registered_groups.json:

"containerConfig": {
  "additionalMounts": [
    { "hostPath": "~/projects/my-app" }
  ]
}

The folder appears inside the container at /workspace/extra/<folder-name> (derived from the last segment of the path). Add "readonly": false for write access, or "containerPath": "custom-name" to override the default name.

8. Configure launchd Service

Generate the plist file with correct paths automatically:

NODE_PATH=$(which node)
PROJECT_PATH=$(pwd)
HOME_PATH=$HOME

cat > ~/Library/LaunchAgents/com.molclaw.plist << EOF
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
    <key>Label</key>
    <string>com.molclaw</string>
    <key>ProgramArguments</key>
    <array>
        <string>${NODE_PATH}</string>
        <string>${PROJECT_PATH}/dist/index.js</string>
    </array>
    <key>WorkingDirectory</key>
    <string>${PROJECT_PATH}</string>
    <key>RunAtLoad</key>
    <true/>
    <key>KeepAlive</key>
    <true/>
    <key>EnvironmentVariables</key>
    <dict>
        <key>PATH</key>
        <string>/usr/local/bin:/usr/bin:/bin:${HOME_PATH}/.local/bin</string>
        <key>HOME</key>
        <string>${HOME_PATH}</string>
    </dict>
    <key>StandardOutPath</key>
    <string>${PROJECT_PATH}/logs/molclaw.log</string>
    <key>StandardErrorPath</key>
    <string>${PROJECT_PATH}/logs/molclaw.error.log</string>
</dict>
</plist>
EOF

echo "Created launchd plist with:"
echo "  Node: ${NODE_PATH}"
echo "  Project: ${PROJECT_PATH}"

Build and start the service:

npm run build
mkdir -p logs
launchctl load ~/Library/LaunchAgents/com.molclaw.plist

Verify it's running:

launchctl list | grep molclaw

9. Test

Tell the user (using the assistant name they configured):

Send @ASSISTANT_NAME hello in your registered chat.

Tip: In your main channel, you don't need the @ prefix — just send hello and the agent will respond.

Check the logs:

tail -f logs/molclaw.log

The user should receive a response in WhatsApp.

Troubleshooting

Service not starting: Check logs/molclaw.error.log

Container agent fails with "Claude Code process exited with code 1":

• Ensure the container runtime is running:
  • Apple Container: container system start
  • Docker: docker info (start Docker Desktop on macOS, or sudo systemctl start docker on Linux)
• Check container logs: cat groups/main/logs/container-*.log | tail -50

No response to messages:

• Verify the trigger pattern matches (e.g., @AssistantName at start of message)
• Main channel doesn't require a prefix — all messages are processed
• Personal/solo chats with requiresTrigger: false also don't need a prefix
• Check that the chat JID is in the database: sqlite3 store/messages.db "SELECT * FROM registered_groups"
• Check logs/molclaw.log for errors

Messages sent but not received by MolClaw (DMs):

• WhatsApp may use LID (Linked Identity) JIDs for DMs instead of phone numbers
• Check logs for Translated LID to phone JID — if missing, the LID isn't being resolved
• The translateJid method in src/channels/whatsapp.ts uses sock.signalRepository.lidMapping.getPNForLID() to resolve LIDs
• Verify the registered JID doesn't have a device suffix (should be [e-mail], not number:[e-mail])

WhatsApp disconnected:

• The service will show a macOS notification
• Run npm run auth to re-authenticate
• Restart the service: launchctl kickstart -k gui/$(id -u)/com.molclaw

Unload service:

launchctl unload ~/Library/LaunchAgents/com.molclaw.plist