Self-Hosting
PocketPaw is designed to be self-hosted. Here’s how to deploy it on your own server.
Requirements
- Python 3.11+
- 2GB RAM minimum (4GB recommended with Mem0)
- Linux, macOS, or Windows (Linux recommended for servers)
- Anthropic API key (for Claude Agent SDK backend)
Basic Deployment
Install
# Create a dedicated user (optional)sudo useradd -m -s /bin/bash pocketpaw
# Install PocketPawcurl -fsSL https://pocketpaw.xyz/install.sh | shConfigure
Create the config file:
mkdir -p ~/.pocketclawcat > ~/.pocketclaw/config.json << 'EOF'{ "anthropic_api_key": "sk-ant-your-key", "agent_backend": "claude_agent_sdk", "tool_profile": "coding"}EOFchmod 600 ~/.pocketclaw/config.jsonRun
pocketpawThe dashboard starts on http://0.0.0.0:8000.
Production Deployment
Behind a Reverse Proxy
For production, put PocketPaw behind nginx or Caddy:
Caddy (recommended):
pocketpaw.example.com { reverse_proxy localhost:8000}nginx:
server { listen 443 ssl; server_name pocketpaw.example.com;
ssl_certificate /etc/ssl/certs/pocketpaw.crt; ssl_certificate_key /etc/ssl/private/pocketpaw.key;
location / { proxy_pass http://localhost:8000; proxy_http_version 1.1; proxy_set_header Upgrade $http_upgrade; proxy_set_header Connection "upgrade"; proxy_set_header Host $host; proxy_set_header X-Real-IP $remote_addr; }}Warning
The WebSocket connection requires Upgrade and Connection headers to be properly proxied. Without these, real-time streaming won’t work.
Environment Variables
Set environment variables in a .env file or your system’s environment:
POCKETCLAW_ANTHROPIC_API_KEY=sk-ant-...POCKETCLAW_DASHBOARD_HOST=127.0.0.1POCKETCLAW_DASHBOARD_PORT=8000POCKETCLAW_TOOL_PROFILE=codingSecurity Hardening
- Run as a dedicated user — Don’t run as root
- Restrict file permissions — Config files should be 600
- Use a firewall — Only expose the reverse proxy port
- Enable Guardian AI — Active by default with Anthropic key
- Set allowed IDs — Restrict channel access to specific users
- Run security audit —
pocketpaw --security-audit
Resource Planning
| Component | RAM | CPU | Notes |
|---|---|---|---|
| PocketPaw core | ~100MB | Minimal | FastAPI + agent loop |
| Mem0 + Qdrant | ~500MB | Minimal | Embedded vector store |
| Ollama (local) | 4-16GB | GPU recommended | Depends on model size |
| Playwright | ~200MB | Moderate | Browser automation |
For a typical deployment with Claude Agent SDK (cloud) and Mem0 with Ollama:
- Minimum: 4GB RAM, 2 CPU cores
- Recommended: 8GB RAM, 4 CPU cores
Was this page helpful?