OpenClaw Integration
Lobstack runs on OpenClaw, the open-source AI agent framework with 215K+ GitHub stars. Learn how the integration works, what version we run, and how to leverage OpenClaw's ecosystem.
What is OpenClaw?#
OpenClaw is the most popular open-source AI agent framework, created by Peter Steinberger. It provides a gateway, tool execution runtime, and messaging platform integrations that power every Lobstack agent.
Lobstack wraps OpenClaw into a managed, one-click deployment platform — no terminal, Docker, or server admin required. You get all the power of OpenClaw with enterprise-grade security, monitoring, and billing.
Version 2026.2.26
Lobstack pins to the latest stable release with all critical security patches (CVE-2026-25253, CVE-2026-24763, CVE-2026-27001).
Security Hardened
Gateway bound to loopback, token auth enforced, sandbox mode enabled, file permissions locked down (chmod 600).
MCP Support
Native Model Context Protocol integration. Connect to 13,000+ MCP servers from the ecosystem.
ClawHub Access
Browse and install from 5,700+ community-built skills on ClawHub, the OpenClaw skill registry.
Architecture#
Lobstack Dashboard -> Lobstack API (Next.js) -> Cloud Provider (Hetzner/DO/Vultr)
|
Dedicated VM (Ubuntu 24.04)
|-- Agent Bridge (Python, port 80)
| |-- REST API (/chat, /config, /skills, /diag)
| |-- WebSocket (port 8765)
| +-- Tool execution engine
|-- OpenClaw Gateway (port 18789)
| |-- Model routing (Anthropic, OpenAI, Google, xAI)
| |-- MCP server orchestration
| +-- Skill & plugin runtime
|-- MCP Servers (filesystem, github, etc.)
+-- Agent config (/root/.openclaw/)Version & Updates#
Lobstack pins OpenClaw to a specific version for reproducibility and security. When a new OpenClaw release is validated, we update the pinned version across all new deployments.
| Property | Value |
|---|---|
| Current Version | 2026.2.26 (February 26, 2026) |
| Minimum Safe Version | 2026.1.29 (CVE-2026-25253 patch) |
| Node.js Requirement | >= 22 |
| Gateway Port | 18789 |
| Install Method | npm install -g openclaw@2026.2.26 |
| Config Location | ~/.openclaw/openclaw.json |
| Skill Directory | ~/.openclaw/skills/ |
Auto-updates
Configuration#
Lobstack generates a secure openclaw.json for each agent with hardened defaults:
{
"gateway": {
"mode": "local",
"bind": "loopback",
"port": 18789,
"auth": { "mode": "token", "token": "<auto-generated>" }
},
"agents": {
"defaults": {
"workspace": "/root/.openclaw/workspace",
"model": "anthropic/claude-sonnet-4-5"
}
},
"security": {
"sandbox": { "mode": "skills-only", "scope": "session", "workspaceAccess": "rw" },
"tools": {
"allowlist": ["run_terminal_command", "read_file", "write_file",
"list_directory", "browse_webpage", "http_request"]
}
},
"mcpServers": {
"filesystem": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-filesystem", "/root/.openclaw/workspace"]
}
}
}MCP Servers#
OpenClaw natively supports the Model Context Protocol (MCP) — an open standard for connecting AI agents to external tools and data sources. MCP servers run as child processes and communicate via JSON-RPC 2.0.
Filesystem MCP
Read, write, and search files in the agent workspace via structured MCP tools.
GitHub MCP
Manage repos, issues, and PRs via the standardized MCP GitHub server.
Web Fetch MCP
Fetch and extract content from URLs with HTML parsing support.
Memory MCP
Persistent key-value storage for cross-conversation context.
ClawHub Skills#
ClawHub is the official OpenClaw skill registry with 5,700+ community-built skills across 40 categories. Skills are Markdown files that teach your agent how to perform specific tasks.
Lobstack agents have the ClawHub CLI pre-installed, allowing skill browsing and installation directly from the agent.
Security notice
Security#
Lobstack applies security hardening beyond OpenClaw defaults:
- Loopback binding — Gateway only listens on 127.0.0.1, never exposed to the internet
- Token authentication — 256-bit random gateway token auto-generated per agent
- Sandbox mode — Enabled for community skills to limit tool execution scope
- File permissions — openclaw.json at chmod 600, credentials at chmod 700
- Security audit —
openclaw security auditruns during provisioning - Version pinning — Specific OpenClaw version deployed for reproducibility
- SSRF protection — IPv6 multicast guards enabled (v2026.2.26)
Patched Vulnerabilities#
| CVE | Severity | Description | Fixed In |
|---|---|---|---|
| CVE-2026-25253 | Critical (8.8) | One-Click RCE via WebSocket Hijack | 2026.1.29 |
| CVE-2026-24763 | High | Docker sandbox bypass | 2026.2.19 |
| CVE-2026-27001 | Medium | Prompt injection via workspace paths | 2026.2.19 |