OpenClaw Security — What You Need to Know Before Giving Your Agent System Access

OpenClaw's power comes from real system access. This guide covers what risks actually exist, what data goes where, and the five security practices that keep your setup safe.

OpenClaw's greatest strength — real access to your file system, terminal, browser and applications — is also the reason security configuration deserves serious attention. An agent that can execute shell commands on your machine has the same access your user account has. Understanding the boundaries of that access and configuring them appropriately is not optional. It is the foundation of a trustworthy setup.

What OpenClaw Accesses and Where Your Data Goes

Your conversation content is sent to your configured LLM provider (OpenAI or Anthropic) if you use cloud models. With Ollama, conversation content stays entirely on your machine. OpenClaw's memory is stored locally in ~/.openclaw/memory/ and never sent to OpenClaw's servers. Your API keys are stored locally in ~/.openclaw/config.yaml and never transmitted. Anonymous usage analytics are collected by default and can be disabled with openclaw config set telemetry false.

The Five Essential Security Practices

1. Protect your configuration file. Your API keys live in ~/.openclaw/config.yaml. Restrict access to owner-only: chmod 600 ~/.openclaw/config.yaml. This prevents other users on the same machine from reading your keys.

2. Review skills before installing. Community skills can request broad system permissions. Before installing any skill from ClawHub, read its SKILL.md to understand what permissions it requests and what commands it runs. The curated awesome-openclaw-skills list filters out known problematic skills.

3. Set API spending limits. In your OpenAI and Anthropic dashboards, set hard monthly spending limits. An unexpected automation loop cannot generate unlimited costs if a ceiling is in place.

4. Use Ollama for sensitive work. Any task involving personal data, confidential business information or anything under regulatory requirements should run with a local Ollama model. Zero data leaves your machine.

5. Keep OpenClaw updated. Security patches are released regularly: npm update -g openclaw. Also update your installed skills: openclaw skills update --all.

Verifying Your Privacy Setup

Run openclaw doctor --privacy at any time to see exactly what data is being sent where in your current configuration. This command shows: which LLM provider your messages are being sent to, whether telemetry is enabled, and which skills have network access. It is the clearest way to understand exactly what your setup is doing.