Frequently Asked Questions

OpenClaw is an open-source AI agent that connects to your messaging apps (WhatsApp, Telegram, Slack, Discord) and runs on your own computer or server.

Think of it as ChatGPT, but:

• You own it completely
• It runs where you want
• It connects to everything you use
• It can automate tasks for you

Learn more about OpenClaw →

Yes, OpenClaw is 100% free and open-source.

However, you'll pay for the AI models you use (like GPT-4 or Claude). These costs typically range from $5-20/month for personal use.

If you host OpenClaw on a cloud server, that's an additional $5-10/month. Running it locally on your own computer is completely free.

No coding required for basic setup!

Installation is done with simple copy-paste commands, and the setup wizard guides you through configuration.

Coding helps if you want to:

• Create custom skills
• Build complex automations
• Contribute to the project

For most users, you'll never need to write code.

About 5-10 minutes for a basic installation.

Breakdown:

• Installing OpenClaw: 2 minutes
• Getting an API key: 2-3 minutes
• Running the setup wizard: 2 minutes
• Your first conversation: instant!

Connecting messaging apps (WhatsApp, Telegram) takes an additional 5 minutes each.

Follow the Quick Start Guide →

Most personal users spend $5-30/month. Here's the breakdown:

AI API Costs (required):

• Light usage: $5-10/month
• Moderate usage: $10-20/month
• Heavy usage: $20-50/month

Hosting (optional):

• Local (your computer): $0
• Cloud VPS: $5-10/month
• Managed (QuickBots): $10-20/month

See the detailed cost guide →

AI models charge per "token" — roughly 4 characters of text. When you chat with OpenClaw, you're charged for both your input and the AI's response.

Typical costs:

• GPT-4o: ~$0.01-0.03 per conversation
• Claude 3 Sonnet: ~$0.01-0.02 per conversation
• GPT-4: ~$0.03-0.10 per conversation

A casual user might have 10-20 conversations per day, costing $0.10-0.60 daily.

Several strategies can significantly reduce your costs:

1. Use cheaper models for simple tasks — Use GPT-3.5 or Claude Haiku for quick questions
2. Set budget limits — OpenClaw can warn you when approaching limits
3. Optimize context — Shorter conversations use fewer tokens
4. Use OpenRouter — Access to many models with one key, easy switching

Read the cost optimization guide →

OpenClaw itself is free, but AI providers typically require payment:

• OpenAI: $5 credit for new accounts
• Anthropic: No free tier
• OpenRouter: Some models have free tiers
• Local models: Completely free (requires good hardware)

You can run smaller local models like Llama 3 for free if you have a decent GPU.

Yes! OpenClaw is designed with security in mind:

• Open source — You can inspect every line of code
• Self-hosted — Your data stays on your machine
• Community audited — Thousands of developers review the code
• No telemetry — No data sent to OpenClaw servers

View the source code →

When self-hosted, only you can see your data.

Your conversations are sent to the AI provider you choose (OpenAI, Anthropic, etc.), subject to their privacy policies. OpenClaw itself doesn't store or transmit your data anywhere.

For maximum privacy, you can:

• Use local models (no external API calls)
• Choose privacy-focused providers
• Review data handling in the code

Only if you explicitly enable it.

By default, OpenClaw can only read files within its workspace directory. To access other files or folders, you need to:

1. Configure file access in settings
2. Grant explicit permissions
3. Approve sensitive operations

OpenClaw always asks before performing potentially destructive actions.

When self-hosted, you control compliance because you control the data.

Considerations:

• Data stays on your infrastructure
• You choose what data to collect/store
• AI provider policies apply to API calls
• No data shared with OpenClaw organization

For business use, consult your legal team about your specific configuration.

OpenClaw supports any model with an OpenAI-compatible API:

• OpenAI: GPT-4, GPT-4o, GPT-3.5
• Anthropic: Claude 3.5 Sonnet, Claude 3 Opus
• Google: Gemini Pro, Gemini Ultra
• OpenRouter: 100+ models from various providers
• Local: Llama, Mistral, Mixtral via Ollama

You can switch models anytime or use different models for different tasks.

Yes! OpenClaw works great with local models via Ollama:

1. Install Ollama
2. Pull a model: ollama pull llama3
3. Configure OpenClaw to use the Ollama endpoint

Benefits of local models:

• No API costs
• Complete privacy
• Works offline

Downsides: Requires good hardware (8GB+ GPU recommended), slower than cloud.

OpenClaw has persistent memory that survives across conversations:

• Short-term: Current conversation context
• Long-term: Facts about you, preferences, past interactions
• Files: Can read/write to workspace files

Memory is stored locally in your OpenClaw configuration folder. You can view, edit, or clear it anytime.

Learn about memory →

OpenClaw is designed to be resilient:

• Gateway auto-restarts if configured with systemd or Docker
• Conversations resume from where you left off
• Memory persists to disk regularly
• No data loss from normal crashes

For maximum uptime, run OpenClaw as a service or in Docker with --restart unless-stopped.

OpenClaw runs on:

• macOS: Intel and Apple Silicon
• Linux: Any distro with Node.js 18+
• Windows: Via WSL2 (recommended) or native
• Docker: Any platform with Docker support
• Raspberry Pi: ARM64 supported

Minimum requirements: 2GB RAM, 1GB disk space.