Open Claw Frequently Asked Questions (FAQ)

Open Claw (OpenClaw) is an open-source autonomous AI personal assistant that runs locally on your devices. It integrates with 11+ messaging platforms and gives AI models "hands and claws" to control your digital life through commands, file operations, and integrations.

Key characteristics:

• Model-agnostic - works with Claude, GPT, local models
• Local deployment - runs on your hardware
• Customizable personality via SOUL.md
• 180K+ GitHub stars with 900+ contributors

Install Open Claw in under 10 minutes with these steps:

• Prerequisites: Node.js 22+, npm, API keys
• Install: npm install -g openclaw@latest
• Configure: openclaw onboard
• Start: openclaw start

See our Installation Guide for detailed platform-specific instructions.

Open Claw has minimal system requirements:

• Node.js: Version 22 or higher (required)
• npm: Package manager included with Node.js
• OS: macOS, Linux, or Windows
• RAM: Can run in <10MB available
• API Keys: For cloud AI models, or local models via Ollama

Model recommendations by use case:

• Coding: Claude 4.5 Opus, GPT-5.3-Codex, qwen3-coder
• General tasks: Claude 4.5 Opus, GPT-5
• Cost-sensitive: Claude 4.5 Haiku, GLM 5, local models
• Privacy-critical: Local models via Ollama
• Chinese language: Kimi K2.5, GLM 5, MiniMax 2.5

See our AI Models Guide for complete details.

Yes, Open Claw software is completely free and open-source. However, you'll need:

• API costs: You pay for AI model API usage (Claude, GPT, etc.)
• Local models: Free if using Ollama with local models
• Your hardware: OpenClaw runs on your own devices

There are no subscription fees for the Open Claw software itself.

SOUL.md is a Markdown file that defines your Open Claw agent's personality. Each workspace has its own SOUL.md file that the agent "reads itself into being" at startup.

You can customize:

• Communication style (formal, casual, witty)
• Values and behavioral boundaries
• Opinions and decision-making approach

See our SOUL.md Guide for templates and examples.

Open Claw has a large skills ecosystem:

• 700+ skills in total ecosystem
• 5,705 community skills on ClawHub (Feb 7, 2026)
• 53 official bundled skills included by default
• 25 core tools built into OpenClaw

Skills extend Open Claw capabilities for productivity, development, communication, and more.

Open Claw supports 11+ messaging platforms:

• WhatsApp, Telegram, Slack, Discord
• Signal, iMessage (BlueBubbles)
• Google Chat, Microsoft Teams
• Matrix, Zalo, WebChat

All platforms can be managed through a single Open Claw agent interface.

Yes, Open Claw supports autonomous operation:

• Periodic check-ins: Agent checks in every 30 minutes to few hours
• Independent task execution: Performs tasks without human intervention
• Proactive notifications: Alerts you to important events
• Scheduled automation: Runs tasks at specified times

See Moltbook experiment for examples of autonomous agent behavior.

Open Claw offers 50+ integrations across categories:

• Chat platforms: All major messaging apps
• Productivity: Email, calendar, task management
• Development: GitHub, GitLab, CI/CD tools
• Smart home: IoT devices, home automation
• Custom: Webhooks for any service

Open Claw is model-agnostic and supports:

• Anthropic: Claude Opus 4.6, Claude 4.5 (Opus, Sonnet, Haiku)
• OpenAI: GPT-5, GPT-5.3-Codex
• Chinese AI: Kimi K2.5, GLM 5, MiniMax 2.5, DeepSeek
• Local models: Via Ollama (qwen3-coder, glm-5, etc.)
• OpenRouter: Access to 100+ models via single API

Configure your AI model in Open Claw:

• Run openclaw onboard to launch configuration wizard
• Select your preferred model provider
• Enter your API key
• Choose default model
• Test with openclaw agent "test message"

Yes, Open Claw supports local models via Ollama (announced Feb 1, 2026):

• Command: ollama launch openclaw
• Requirement: Models with ≥64,000 token context length
• Recommended: qwen3-coder, glm-5, gpt-oss:20b, kimi-k2.5
• Benefits: Zero API costs, complete privacy, offline operation

OpenRouter is a unified API gateway built into Open Claw:

• Auto Model: Automatically selects best model for each task
• Cost optimization: Routes to most cost-effective provider
• Single API: Access 100+ models with one key
• Fallback: Switches providers if one fails

For coding tasks, we recommend:

• GPT-5.3-Codex: Most capable agentic coding model
• Claude 4.5 Opus/Sonnet: Excellent reasoning for complex code
• qwen3-coder (local): Best free offline option

All provide strong performance for code generation, debugging, and technical tasks.

Open Claw requires careful configuration and security awareness:

• Patched vulnerabilities: Critical CVEs have been fixed
• VirusTotal integration: Skills scanned automatically (v2026.2.6)
• User responsibility: Must follow security best practices
• Early-stage software: New vulnerabilities may emerge

See our Security Guide for detailed information.

Key security concerns with Open Claw:

• CVE-2026-25253: Critical RCE vulnerability (patched in v2026.1.29)
• Malicious skills: 341 malicious skills identified on ClawHub
• Secret leaks: 7.1% of skills mishandle API keys
• Broad permissions: Agent has extensive system access

Always keep OpenClaw updated and review skills before installation.

Follow these security best practices:

• Always update to latest version
• Never expose web interface (127.0.0.1:18789) publicly
• Only install VirusTotal-scanned skills
• Review skill permissions before installation
• Use dedicated API keys with limited scopes
• Run on isolated network for sensitive data

Exercise caution with community skills:

• 341 malicious skills identified on ClawHub
• 7.1% of skills mishandle secrets (API keys, credit cards)
• VirusTotal scanning now integrated (v2026.2.6)
• Always review skill source code before installing

Only install skills from trusted sources that have been scanned by VirusTotal.

Open Claw requires broad permissions to function:

• Email accounts (read, send, manage)
• Calendar data (view, modify)
• Messaging platforms (all connected accounts)
• File systems (read, write, execute)
• Shell commands (execute arbitrary commands)
• Connected services and APIs

These permissions mean a compromised instance gives attackers extensive access.

Open Claw software pricing:

• Software: Completely free and open-source
• No subscription fees: Zero recurring costs
• BYOK: Bring your own API keys
• Local models: Free via Ollama integration

API costs depend on your chosen model and usage:

• Claude Opus 4.6: Highest cost, best capability
• Claude 4.5 Haiku: Lowest cost, fastest responses
• GPT-5: Moderate cost, strong performance
• Local models: Zero API costs (requires hardware)

Monitor token usage via Open Claw Web UI dashboard (v2026.2.6).

Cost optimization strategies:

• Use OpenRouter Auto for automatic optimization
• Match model to task complexity (Haiku for simple, Opus for complex)
• Switch to local models for high-volume, simple tasks
• Enable caching to reduce duplicate API calls
• Set budget limits in provider dashboards

Yes, you can use Open Claw completely free:

• Local models via Ollama: Zero API costs after setup
• Recommended models: qwen3-coder, glm-5, gpt-oss:20b
• Requirements: ≥64K token context, sufficient hardware
• Trade-offs: May have lower capability than cloud models

Join the Open Claw community through:

• Discord: discord.com/invite/clawd
• GitHub: github.com/openclaw/openclaw
• X/Twitter: Active community discussions
• Documentation: docs.openclaw.ai

180K+ GitHub stars and 900+ contributors are already part of the community.

Ways to contribute to Open Claw:

• Submit pull requests on GitHub
• Create and share skills on ClawHub
• Report bugs and feature requests
• Improve documentation and tutorials
• Help other users on Discord
• Translate docs to other languages

Get help through these channels:

• Discord #help: Real-time community support
• GitHub Issues: Bug reports and feature requests
• Documentation: Comprehensive guides at docs.openclaw.ai
• FAQ: This page for common questions
• Stack Overflow: Tagged technical questions

Report bugs through:

• GitHub Issues: github.com/openclaw/openclaw/issues
• Security issues: Report responsibly through proper channels
• Discord: #bugs channel for discussion

Include version number, steps to reproduce, and expected vs actual behavior.