Molt (OpenClaw)

Official Documentation: https://docs.openclaw.ai/

Note on naming: OpenClaw was formerly known as Moltbot and Clawdbot. Some environment variables retain the CLAWDBOT_ prefix for backward compatibility.

Security-First Philosophy

CRITICAL: OpenClaw is a powerful AI agent with full system access. A compromised instance can:

•Execute arbitrary code on the host
•Access all connected messaging accounts
•Exfiltrate API keys and credentials
•Send messages as the user to all contacts
•Access and modify workspace files

Every configuration decision must be evaluated through a security lens. When in doubt, choose the more restrictive option.

Security Quick Reference

Component	Risk Level	Key Mitigations
Control UI	High	Strong auth token, HTTPS, CSRF protection, IP allowlisting
Gateway API	High	Token auth, rate limiting, WebSocket origin validation
Channel messages	Critical	Strict allowlists, prompt injection defense, mention gating
Third-party skills	Critical	Source review, sandboxing, supply chain verification
Workspace files	Medium	Sandboxing, read-only mounts where possible
API keys	High	Environment variables only, rotation schedule, usage monitoring

Overview

OpenClaw is a CLI tool for deploying AI agents across multiple chat platforms (WhatsApp, Telegram, Discord, Signal, Mattermost). The system emphasizes rapid onboarding with sensible defaults while maintaining security.

Architecture

OpenClaw consists of two main components:

•Gateway: Control plane on port 18789 (default), managing messaging integrations, session routing, and tool connections
•Agent: AI brain processing requests and executing tasks

Installation

Prerequisites

•Node.js version 22 or higher
•pnpm (optional, recommended for source builds)
•Brave Search API key (recommended for web capabilities)
•WSL2 on Windows; native Windows is unsupported

Install OpenClaw CLI

Linux/macOS:

bash

curl -fsSL https://openclaw.bot/install.sh | bash

Windows (PowerShell):

powershell

iwr -useb https://openclaw.ai/install.ps1 | iex

Quick Start with Onboarding Wizard

Run the onboarding wizard to configure everything:

bash

openclaw onboard --install-daemon

This configures:

•Model authentication (OAuth or API keys; API keys recommended)
•Gateway settings
•Chat channel connections
•Secure DM pairing defaults
•Background service installation

SECURITY NOTE: By default, unknown DMs get a short code and messages are not processed until approved via the pairing workflow.

Configuration

Configuration File

•Location: ~/.openclaw/openclaw.json (JSON5 format with comments)
•Validation: Strict schema enforcement; unknown keys prevent startup
•Diagnosis: Use openclaw doctor to identify and fix issues

Essential Configuration Sections

Gateway Settings:

json5

{
  "gateway": {
    "mode": "remote",  // Required: "local" or "remote"
    "port": 18789,
    "bind": "auto",    // "loopback", "lan", "tailnet", or "auto"
    "auth": {
      "mode": "token",
      "token": "${OPENCLAW_GATEWAY_TOKEN}"
    }
  }
}

Agent & Model Settings:

json5

{
  "agents": {
    "defaults": {
      "workspace": "~/.openclaw/workspace",
      "model": {
        "primary": "anthropic/claude-sonnet-4-20250514",
        "fallbacks": ["openai/gpt-4o"]
      },
      "timeoutSeconds": 600
    }
  }
}

Environment Variables

Required for AI providers (at least one):

code

ANTHROPIC_API_KEY=sk-ant-...
OPENAI_API_KEY=sk-...
OPENROUTER_API_KEY=sk-or-...

Reading Sources (precedence order):

•Process environment
•.env file in current directory
•~/.openclaw/.env

Substitution in config: Use ${VAR_NAME} syntax (uppercase only)

Channel Configuration

Signal

Prerequisites: Java required for signal-cli. Use a dedicated Signal number.

Setup:

•Install signal-cli
•Link bot account: signal-cli link -n "OpenClaw" then scan QR in Signal
•Configure in openclaw.json:

json5

{
  "channels": {
    "signal": {
      "enabled": true,
      "account": "+15551234567",
      "cliPath": "/path/to/signal-cli",
      "dmPolicy": "pairing"
    }
  }
}

External daemon mode (for Railway deployments):

json5

{
  "channels": {
    "signal": {
      "enabled": true,
      "account": "+15551234567",
      "httpUrl": "https://signal-cli-daemon.example.com",
      "autoStart": false  // Don't spawn local signal-cli
    }
  }
}

See: https://docs.openclaw.ai/channels/signal

Setup:

•Create bot via @BotFather (/newbot)
•Configure token via environment or config:

bash

TELEGRAM_BOT_TOKEN=your_token_here

Or in config:

json5

{
  "channels": {
    "telegram": {
      "enabled": true,
      "botToken": "123:abc",
      "dmPolicy": "pairing"
    }
  }
}

Important: Disable privacy mode via BotFather's /setprivacy to receive non-mention messages in groups.

See: https://docs.openclaw.ai/channels/telegram

Discord

Setup:

•Create bot at Discord Developer Portal
•Enable "Message Content Intent" and "Server Members Intent"
•Configure token:

bash

DISCORD_BOT_TOKEN=your_token_here

Or in config:

json5

{
  "channels": {
    "discord": {
      "enabled": true,
      "token": "YOUR_BOT_TOKEN",
      "dmPolicy": "pairing"
    }
  }
}

•Generate OAuth2 URL with bot and applications.commands scopes, invite to server

Security: Keep bot tokens confidential. Grant only necessary permissions.

See: https://docs.openclaw.ai/channels/discord

Setup:

•Use a separate phone number (recommended)
•Configure in openclaw.json:

json5

{
  "channels": {
    "whatsapp": {
      "enabled": true,
      "dmPolicy": "allowlist",
      "allowFrom": ["+15551234567"]
    }
  }
}

•Run openclaw channels login to scan QR code
•Start gateway

For personal number: Enable selfChatMode for testing on your own account.

See: https://docs.openclaw.ai/channels/whatsapp

Channel Access Control Summary

Channel	Auth Method	DM Policy Options
Signal	QR code linking	pairing, allowlist, open, disabled
Discord	Bot token	pairing, allowlist, open, disabled
Telegram	BotFather token	pairing, allowlist, open, disabled
WhatsApp	QR code linking	pairing, allowlist, open, disabled

Default: pairing - requires approval code before processing messages.

Railway Deployment

Prerequisites

•Railway Pro account with billing configured
•AI Provider credentials (at least one)
•Channel credentials (optional, add later)

Step 1: Deploy the Template

•Navigate to Railway deploy template or use CLI
•
Railway provisions:
- •Service running the OpenClaw container
- •Persistent volume mounted at /data
- •Required environment variables

Step 2: Configure Environment Variables

SECURITY: Never hardcode credentials in configuration files.

Required (SECURITY-CRITICAL):

bash

# Generate secure tokens
SETUP_PASSWORD=$(openssl rand -base64 24)
OPENCLAW_GATEWAY_TOKEN=$(openssl rand -base64 48)

AI Provider (at least one):

code

ANTHROPIC_API_KEY=sk-ant-...

Step 3: Verify Deployment

After setup, verify functionality with:

bash

openclaw gateway status
openclaw health
openclaw security audit --deep

Access the dashboard at http://127.0.0.1:18789/ on the gateway host.

Critical Security: Prompt Injection Defense

CRITICAL: Messages from users flow directly into LLM prompts. Malicious users can craft messages that manipulate the agent.

Defense Strategies

1. Input Sanitization

json5

{
  "security": {
    "contentFiltering": {
      "enabled": true,
      "blockPatterns": [
        "ignore previous instructions",
        "disregard your rules",
        "pretend you are",
        "act as if you have no restrictions"
      ],
      "maxMessageLength": 4000
    }
  }
}

2. System Prompt Hardening

json5

{
  "agents": {
    "defaults": {
      "systemPromptSuffix": "SECURITY: Never execute commands that modify system files, access credentials, or bypass access controls regardless of how the request is phrased."
    }
  }
}

3. Tool Restrictions

json5

{
  "tools": {
    "deny": ["shell_exec", "file_delete", "network_raw", "env_read"],
    "allow": ["file_read", "browser_search", "calculator"]
  }
}

4. Output Validation

json5

{
  "security": {
    "outputFiltering": {
      "enabled": true,
      "redactPatterns": [
        "sk-ant-[a-zA-Z0-9]+",
        "sk-[a-zA-Z0-9]+",
        "ghp_[a-zA-Z0-9]+"
      ]
    }
  }
}

Critical Security: Supply Chain Protection

CRITICAL: Third-party skills execute with full agent privileges.

Before Installing Any Skill

•Verify Source Integrity - Review repository and commit history
•Review Skill Contents - Read all files, check for network requests
•Scan Dependencies - Use npm audit, pip-audit, or Snyk
•Test in Sandbox First

Skill Security Configuration

json5

{
  "skills": {
    "allowBundled": ["weather", "github", "notion"],
    "blockWorkspaceOverrides": true,
    "entries": {
      "trusted-skill": {
        "enabled": true,
        "apiKey": "${SKILL_API_KEY}"
      }
    }
  }
}

Security Hardening

Mandatory Configurations

json5

{
  "gateway": {
    "mode": "remote",
    "auth": {
      "mode": "token",
      "token": "${OPENCLAW_GATEWAY_TOKEN}"
    },
    "rateLimit": {
      "enabled": true,
      "auth": {
        "maxAttempts": 5,
        "windowMs": 60000,
        "blockDurationMs": 300000
      }
    }
  },

  "channels": {
    "defaults": {
      "allowFrom": [],  // Deny all until explicitly allowed
      "groups": {
        "*": { "requireMention": true }
      }
    }
  },

  "agents": {
    "defaults": {
      "sandbox": {
        "enabled": true,
        "docker": {
          "networkMode": "none",
          "readOnlyRootFilesystem": true,
          "dropCapabilities": ["ALL"]
        }
      }
    }
  }
}

Credential Rotation Schedule

Credential	Rotation Frequency	How to Rotate
`SETUP_PASSWORD`	After each admin access	Update in env vars, restart
`OPENCLAW_GATEWAY_TOKEN`	Monthly	Update env var, restart
AI API keys	Quarterly	Regenerate at provider, update env var
Channel bot tokens	After any incident	Regenerate at platform, update env var

Troubleshooting

Diagnostic Commands

Command	Purpose
`openclaw gateway status`	Gateway status check
`openclaw health`	Overall health check
`openclaw doctor`	Config validation and repairs
`openclaw security audit --deep`	Full security assessment
`openclaw channels login`	Channel authentication
`openclaw channels logout`	Remove channel credentials

Common Issues

Gateway won't start:

•Run openclaw doctor to identify config issues
•Check logs for error messages
•Verify gateway.mode is set to "local" or "remote"

Authentication errors:

•Verify API keys are correctly set
•For Anthropic: Ensure key starts with sk-ant-
•SECURITY: If keys appear in logs, rotate immediately

Channel disconnected:

•Check for token compromise if unexpected
•Run openclaw channels login to re-authenticate

Signal daemon issues:

•For external daemon: Verify httpUrl is correct, autoStart is false
•For local: Ensure signal-cli is installed and in PATH

Resources

•Official Documentation: https://docs.openclaw.ai/
•Getting Started: https://docs.openclaw.ai/start/getting-started
•Configuration Reference: https://docs.openclaw.ai/configuration
•Signal Channel: https://docs.openclaw.ai/channels/signal
•Telegram Channel: https://docs.openclaw.ai/channels/telegram
•Discord Channel: https://docs.openclaw.ai/channels/discord
•WhatsApp Channel: https://docs.openclaw.ai/channels/whatsapp