OpenClaw Setup
& Custom Claude Code Deployment

OpenClaw (formerly known as ClawdBot and MoltBot) is an open-source personal AI assistant created by Peter Steinberger. It runs on your own hardware or a cloud server, connects to messaging platforms like Telegram, WhatsApp, Discord, Slack, Signal, and iMessage, and can browse the web, execute code, manage files, send emails, and automate tasks. Unlike cloud-hosted AI assistants, OpenClaw gives you full control over your data because everything runs on your own infrastructure. The project has grown to over 430,000 lines of code and supports multiple AI model providers including Claude, GPT, Gemini, Grok, and local models via Ollama. It is essentially a full-time AI assistant that works 24/7 on your behalf, accessible from any messaging platform you already use daily.

OpenClaw is a general-purpose AI assistant framework with around 430,000 lines of code. Our custom Claude Code setup uses a completely different architecture: Anthropic's official Claude Code CLI enhanced with 267 specialized AI agents, 190+ custom skills, multi-agent orchestration via our AI Super Brain (AISB), automated content pipelines, security scanning tools, and enterprise-grade VPS infrastructure. The practical difference is depth versus breadth. OpenClaw connects to many platforms and handles general tasks well — web browsing, messaging, basic code execution, file management. Our Claude Code setup goes deep on software development, DevOps, security, marketing automation, and multi-agent coordination. For a detailed breakdown of the Claude Code system, see our Claude Code setup service. Think of OpenClaw as a Swiss Army knife and our Claude Code setup as a fully equipped workshop — both get the job done, but one handles significantly more complex work.

OpenClaw itself is free and open-source (MIT license). Your costs are hosting ($4-30/month for a VPS) plus AI model API usage ($1-150/month depending on how much you use it). You can even run OpenClaw for free using Oracle Cloud's Always Free tier with 4 ARM CPUs and 24GB RAM. Our professional setup starts from EUR 500 for standard OpenClaw installation with messaging integrations and security hardening. The custom Claude Code setup with 267 agents, AISB orchestration, and full enterprise configuration ranges from EUR 1,500 to 3,000. For most users, the total monthly cost ranges from $5-50 for hosting plus API usage, which is a fraction of what a human assistant or SaaS tool subscription would cost for comparable capabilities.

Setting up OpenClaw with Telegram involves creating a bot via @BotFather, configuring the bot token in OpenClaw's settings, and setting up the gateway. While the basic steps take a few minutes, proper configuration requires attention to detail: privacy mode settings to control who can interact with your bot, group policies for multi-user environments, inline command registration for quick actions, error handling and reconnection logic for reliability, and rate limiting to prevent API abuse. We handle the entire Telegram integration as part of our setup service, including testing every command type, verifying file sharing works correctly, and ensuring the bot restarts automatically if the server reboots. The result is a Telegram bot that responds instantly, handles media, and works reliably 24/7.

Yes. OpenClaw connects to WhatsApp through the WhatsApp Web protocol. During setup, OpenClaw displays a QR code that you scan with your WhatsApp app, similar to logging into WhatsApp Web. We configure the full WhatsApp integration including media support, group handling, and reconnection logic. WhatsApp is one of the most popular messaging platforms worldwide, and having an AI assistant accessible through it means you can send tasks, ask questions, and receive results from the same app you use for daily communication. The integration supports text messages, voice notes, images, documents, and group chats.

OpenClaw supports Claude (Anthropic — requires API key since January 2026 after OAuth was deprecated), GPT-4o and GPT-5 (OpenAI), Gemini (Google), Grok (xAI), DeepSeek, Kimi, MiniMax, Qwen, and local models via Ollama. You can use BYOK (Bring Your Own Key) with zero markup, and route different tasks to different models based on cost or capability.

We can provision a VPS for you (Hetzner from $4/month, DigitalOcean from $6/month, AWS Lightsail from $5/month, or Oracle Cloud for free) or work with hardware you already own — Mac Mini, Linux PC, even a Raspberry Pi. For always-on reliability, we recommend a cloud VPS. We handle all server setup, SSH hardening, and security configuration.

Standard OpenClaw setup takes under 1 hour from start to finish, including messaging integrations and testing. Our custom Claude Code setup takes 2-4 hours depending on the number of integrations, custom agents, and automation pipelines. Both include full verification — nothing is handed off until everything works. This is dramatically faster than self-setup, which typically takes 2-3 days of trial and error for OpenClaw and up to a week for a full Claude Code environment with all integrations. The time savings come from our experience with dozens of identical deployments and pre-built configuration templates that we customize to your specific needs.

Yes. Everything runs on your own dedicated server — not shared infrastructure. Your conversations, files, code, and data never leave your machine unless you explicitly send them to an AI model provider via API. With BYOK, you have direct relationships with AI providers. We configure SSH key-only authentication, firewalls, fail2ban, and non-root user access as part of every setup.

OpenClawInstall.AI offers managed OpenClaw hosting starting at $29/month with automated provisioning. We offer a different value: professional one-time setup with security hardening, custom integrations, hands-on configuration, and the option to upgrade to our full Claude Code system with 267 agents and 190+ skills. If you want ongoing managed hosting, they are a good option. If you want a one-time professional setup on your own server with more power, we are the better choice.

Nanobot is a lightweight Python alternative (~4,000 lines vs OpenClaw's 430,000+). NanoClaw focuses on container isolation for security. Both are simpler but have smaller ecosystems. Our custom Claude Code setup takes a different approach entirely — instead of trimming OpenClaw down, we built a more capable system from the ground up with Anthropic's official tooling, 267 specialized agents, and production-tested workflows.

Yes. OpenClaw runs on macOS, Linux, and Windows (via WSL2). We have deployed on Mac Minis, dedicated servers, Raspberry Pi 4/5 devices, and various cloud platforms. For always-on reliability, a cloud VPS is recommended, but local hardware works well for personal use. We configure the daemon to run at boot so your agent restarts automatically.

MCP (Model Context Protocol) servers extend OpenClaw's capabilities by connecting it to external tools and services — GitHub, Gmail, Google Calendar, Stripe, databases, APIs, and more. We configure the MCP servers most relevant to your workflow during setup, enabling your AI agent to interact with your existing tools directly. For a comprehensive MCP configuration with custom server development, see our dedicated MCP setup service.

Yes. OpenClaw supports local models through Ollama, which means you can run AI models entirely on your own hardware with zero API costs and complete privacy. This works best on machines with at least 16GB RAM. We can configure a hybrid setup where local models handle simple tasks and cloud models (Claude, GPT) handle complex ones.

OpenClaw was previously known as ClawdBot and before that as MoltBot. The project was rebranded to OpenClaw in late 2025. All the setup guides and tutorials for ClawdBot and MoltBot apply to OpenClaw — it is the same project with a new name and continued active development.

You can install OpenClaw yourself in about 10 minutes with the one-liner curl command. But getting it production-ready — SSH hardening, Telegram bot with proper privacy settings, MCP servers, custom skills, persistent memory optimization, web search providers, model routing, daemon configuration, and automation pipelines — takes days of trial and error. We have done this dozens of times and our setup is battle-tested across real production environments with over 267 agents and 190 skills. You skip the learning curve entirely. For an even more powerful system, see our Claude Code setup service with full agent orchestration, or pair it with our Cursor IDE configuration for the ultimate development environment.