Open-source self-hosted personal agent & ClawHub host Fact-checked · 2026-11-17

OpenClaw Hub: Self-Hosting, Chat Channels, Security Risks and ClawHub Plugins

Understand the open-source personal agent OpenClaw: model costs, Telegram/Discord/WhatsApp gateways, why you should isolate its shell execution in containers, and catalogued ClawHub plugins.

Key takeaways

  • OpenClaw (`openclaw/openclaw`) is an MIT-licensed self-hosted personal agent controlled via Telegram, Discord, WhatsApp or web dashboards.
  • While the software is free, autonomous multi-step loops and memory summarization can consume significant API tokens—set hard spending limits on your provider.
  • Security warning: OpenClaw can execute local shell commands and drive browsers; run it inside a container and audit third-party ClawHub skills before installing.
  • Supported by a wide ecosystem of channel bridges, management dashboards and skill packs.

Key facts: pricing, models, surfaces & safety

Pricing & quota
MIT open-source free · Bring your own API keys or local models

Free to self-host. You pay only your model provider (or run local Ollama models); configure provider spending caps to prevent runaway loop costs.

Models & providers
Claude, OpenAI, Gemini, DeepSeek, OpenRouter & local Ollama

Supports major cloud LLM APIs and local inference servers.

OS & surfaces
macOS · Linux · Windows (WSL/Docker) · Chat app gateways

Runs as a persistent daemon reachable from messaging apps and a local web UI.

Safety & permissions
High local privileges · Isolate in Docker & never expose unauthenticated ports

Because untrusted messages, emails or web pages can trigger indirect prompt injection, isolate OpenClaw in a container and restrict host filesystem mounts.

Extension points: ClawHub Skills, Chat Gateways and Dashboards

Community projects extend OpenClaw across three layers:

ClawHub & OpenClaw Skills

Installable task skills and workflow automations—review their source before granting shell or secret access.

Chat Channel Bridges

Connectors for Telegram, Discord, WhatsApp, Slack and regional messaging platforms.

Management & Guardrail Tools

Dashboards, memory layers and safety wrappers for persistent OpenClaw instances.

OpenClaw ecosystem & extensions

Plugins, config packs, switchers and orchestration layers in the directory built for OpenClaw, by stars; each has its own FARS review.

Plugins, routing & companion extensions

Automation & Ops

1Panel

Manage Linux servers, containers, websites, and AI agents from one web interface.

Insufficient evidence ★ 37k · 1d ago
Automation & Ops ✓ NVIDIA · Official

NVIDIA NemoClaw

A secure reference stack for operating always-on agents in NVIDIA OpenShell sandboxes.

81 · Good ★ 23k · today
Automation & Ops

Edict Three-Departments Orchestrator

An OpenClaw workflow that plans, reviews, dispatches, and tracks multi-agent work through named institutional roles.

45 · Major gaps ★ 17k · 1mo ago
Data & Analysis

AutoResearchClaw

Turns a research topic into experiments, verified citations, and paper-ready LaTeX deliverables.

58 · Major gaps ★ 15k · 3mo ago
Writing & Content

InkOS Story Creation Workspace

A stateful writing agent system for novels, scripts, interactive fiction, and multilingual book translation.

71 · Some gaps ★ 11k · 28d ago
Automation & Ops

AI-Infra-Guard (A.I.G) – AI Red Teaming Platform by Tencent Zhuque Lab

Full-stack AI security self-inspection: OpenClaw security scan, agent/MCP/skill scan, AI infra vulnerability scan, and jailbreak evaluation.

49 · Major gaps ★ 7.6k · 1d ago
Productivity & Collaboration

ClawX Desktop

A no-terminal desktop workspace for configuring, chatting with, and automating OpenClaw AI agents.

59 · Major gaps ★ 7.6k · 1d ago
Dev & Engineering

AI Coding & Agent Workflows in Chinese: A Practical Guide

A systematic Chinese tutorial collection covering Claude Code, OpenClaw, Codex, and WorkBuddy, helping individuals, teams, and institutions adopt AI coding and agent workflows.

64 · Some gaps ★ 6.5k · 11d ago
Productivity & Collaboration

LobsterAI

An open-source desktop assistant for executing office work in your real local environment.

60 · Some gaps ★ 6.2k · 1d ago
Automation & Ops

SkillHub

A private enterprise registry for publishing, governing, and distributing agent skill packages.

62 · Some gaps ★ 5.4k · 1d ago
Data & Analysis

MetaClaw

A local proxy that turns personal-agent conversations into skills, memory, and optional scheduled LoRA learning.

49 · Major gaps ★ 3.5k · 5mo ago
Automation & Ops

Narrator AI CLI Skill

Teach your AI agent to create movie narration videos

59 · Major gaps ★ 3.3k · 4mo ago
Productivity & Collaboration

Nexu Desktop

A local desktop client that brings an OpenClaw Agent into WeChat, Feishu, Slack, and Discord.

57 · Major gaps ★ 3.3k · 7mo ago
Productivity & Collaboration

TuriX

A configurable desktop-use agent that carries out GUI tasks across browser and office workflows.

41 · Major gaps ★ 3.2k · 2mo ago
Automation & Ops

ClawPanel

A cross-platform control panel for installing, configuring, diagnosing, and operating OpenClaw and Hermes Agent.

48 · Major gaps ★ 2.9k · 3d ago
Automation & Ops

SkillClaw: Collective Skill Evolution Framework

Let AI agents evolve skills from real interactions—across sessions, agents, devices, and users. Install and forget; evolution happens in the background.

49 · Major gaps ★ 2.8k · 1mo ago
Automation & Ops

OpenClaw Master Skills

A curated collection of 1209+ best OpenClaw skills, updated weekly by MyClaw.ai.

42 · Major gaps ★ 2.2k · 2d ago
Productivity & Collaboration

Team9 - Collaborative Workspace for AI Agents

A collaborative workspace for AI agents built on OpenClaw, enabling instant messaging, multi-agent management, and document sharing.

42 · Major gaps ★ 2k · 5mo ago

Browse OpenClaw skills on FollowSkills →

OpenClaw comparisons & rankings

Related concepts

Frequently asked questions

Should I choose OpenClaw or Hermes Agent?

Both are MIT-licensed self-hosted personal agents. OpenClaw has a larger community plugin catalog (ClawHub), while Nous Research's Hermes Agent ships with clearer command-approval defaults (`smart` mode), Docker/Modal execution backends and structured skill self-improvement.

What are the main security risks when running OpenClaw?

Running directly on your host OS with full home-directory access, exposing the gateway port to the public internet, or installing unverified third-party skills. Always use container isolation and strict chat allowlists.

Host agent ecosystem hubs