OpenClaw Hardened Deployment (Ansible)

Automated Tier 3+ Security Hardening for OpenClaw AI Agents

This Ansible playbook implements and extends the security hardening measures described in the OpenClaw Security Guide, providing a fully automated deployment with additional defense-in-depth layers.

🎯 What This Playbook Does

Deploys a hardened OpenClaw installation with:

• Rootless Podman containers: Strict isolation running as a non-privileged user.
• Network egress filtering: Squid proxy sidecar with a domain allowlist.
• HTTPS termination: Caddy reverse proxy with auto-generated self-signed certificates.
• LiteLLM credential brokering: OpenClaw never sees real API keys; LiteLLM spoofs models (e.g., Deepseek acting as Claude).
• Consolidated Configuration: Single openclaw.json master config for Gateway, Tools, and Agents.
• Automated Identity: EFF wordlist hostname generation and persistent SSH key management.
• Multi-OS support: Native tasks for Arch Linux and Debian/Ubuntu (AWS ready).
• Security Monitoring: Systemd-based weekly audits for prompt injections and blocked domains.

📊 Comparison: Article vs. This Implementation

Feature	Original Article (Tier 3)	This Ansible Implementation
Container Runtime	Docker	Podman (rootless) ⭐
Network Filtering	Firewall only	Firewall + Squid egress allowlist ⭐
HTTPS	Optional/Manual	Caddy reverse proxy (Terminated HTTPS) ⭐
Identity Management	Manual setup	Automated EFF wordlist generation
OS Support	Ubuntu focus	Arch + Debian/Ubuntu auto-detection
Deployment Method	Manual	Fully automated interactive script
Monitoring	Manual cron	Systemd timers + audit script
LLM Providers	Anthropic focus	Ollama (Deepseek) / Anthropic / OpenAI
Secrets Management	Manual generation	Auto-gen with PERSISTENCE across runs ⭐
Access Control	Token only	Token + Manual Device Pairing

📋 Prerequisites

Local Machine (Controller):

• Ansible 2.10+
• OpenSSL (for cert generation)
• SSH Client (ssh-keygen)
• Python 3.8+

Target Machine:

• Arch Linux OR Debian/Ubuntu
• Initial root/sudo access (Password or AWS .pem key)
• 2GB+ RAM

🚀 Quick Start

1. Prepare

cd openclaw-hardened-ansible
chmod +x deploy.sh update-allowlist.sh

2. Deploy

Run the interactive script. It will prompt for your IP, provider, and keys.

./deploy.sh

AWS/Cloud Example:

./deploy.sh \
  --target 54.x.x.x \
  --ssh-user ubuntu \
  --ssh-key ~/my-aws-key.pem \
  --mgmt-cidr 192.168.20.0/24 \
  --provider ollama \
  --model "deepseek-r1:8b" \
  --url "http://10.100.1.25:11434"

3. Authenticate

Once finished, get your persistent token:

ssh -i ssh-keys/your-name.pem openclaw@IP "cat ~/openclaw-docker/.env | grep TOKEN"

Open https://IP:18789, click through the SSL warning, and paste the token in Settings.

4. Approve Device (Hardening Step 7)

Since device auth is enabled, you must approve your browser from the host CLI:

# Inside the OpenClaw host
podman exec openclaw-agent openclaw devices pending
podman exec openclaw-agent openclaw devices approve <YOUR_ID>

🔧 Maintenance

Update Egress Allowlist

1. Edit roles/tier3-setup/templates/allowlist.txt.j2.
2. Run ./update-allowlist.sh -t IP --ssh-user USER --ask-pass.

Security Audits

A systemd timer runs monitor-openclaw.sh weekly. To run manually:

sudo /home/openclaw/openclaw-docker/monitor-openclaw.sh

Check the reports at ~/openclaw-docker/security-audit-YYYYMMDD.log.

Configuration Validation

If you see errors, run the OpenClaw "Doctor" to check the schema:

podman exec openclaw-agent openclaw doctor

📁 File Structure

• deploy.sh: Main entry point (interactive/CLI).
• update-allowlist.sh: Lightweight allowlist updater.
• ssh-keys/: Stores generated .pem and .crt files.
• roles/tier3-setup/: The core hardening logic.
• requirements.yml: Ansible dependencies (auto-installed).

📄 License

Provided as-is for harm-reduction. OpenClaw is architecturally "spicy"—this deployment reduces the blast radius but prompt injection remains an inherent risk of LLMs. Use burner accounts only.