OpenClaw on DeepLayer
How OpenClaw runs in the DeepLayer environment.
Architecture
DeepLayer runs one OpenClaw gateway per customer instance. This follows the OpenClaw recommended deployment model where each trust boundary gets its own gateway process with isolated state.
What we manage
- Gateway process lifecycle (start, stop, restart)
- Persistent storage at
~/.openclawfor config, sessions, and agent data - Reverse proxy with TLS termination
- Automated backups of the full OpenClaw state directory
What you control
- Agent configuration and channel connections
- Model provider settings (API keys, base URLs)
- Workspace files and agent behavior
Storage layout
Your instance data is stored under a dedicated volume:
~/.openclaw/
├── agents/ # Per-agent config, sessions, auth
├── config/ # Gateway configuration
└── workspace/ # Workspace files
Supported OpenClaw features
All standard OpenClaw features work on DeepLayer, including:
- Multi-agent setups
- Channel connections (with appropriate API keys)
- Workspace and file management
- Session transcripts and history
Limitations
- Direct SSH access to the underlying host is not available
- Custom gateway plugins are not supported in v1
- Outbound network access from sandboxed workspaces may be restricted