UnofficialGuide
officialSite

OpenClaw Guide for Everyone

What is OpenClaw, GitHub history, install, Docker, Telegram, QQ, Ollama, skills, security, cost, and Claude Code comparisons.

Primer

What Is OpenClaw?

A self-hosted AI agent system that acts more like an operator than a chat app. OpenClaw is free and open source, built to run across channels, tools, and memory, and powerful enough that security and cost control belong in the first conversation.

Self-hosted AI agent, not just a chat box

OpenClaw connects channels, runs tools, stores memory, and can act through a long-running gateway instead of a single hosted chat tab.

20+ channels, 55 bundled skills, 13,729 registry entries

The ecosystem is already large, but quality and safety vary widely, especially in the wider skills market.

Security and cost are first-order concerns

CVE-2026-25253, malicious skills, exposed gateways, and runaway token spend are all part of the real OpenClaw story in 2026.

China deployment and IM adoption are major growth drivers

QQ, Feishu, DingTalk, Tencent Cloud, Aliyun, and Volcengine are central to the Chinese OpenClaw ecosystem.

Channels

Start with the easiest channel, not the most impressive one

Telegram is the cleanest zero-public-IP route. QQ and Feishu matter for China. Discord and WhatsApp cover community and daily messaging. Remote access is its own layer.

Models

Model freedom is one of OpenClaw’s biggest advantages

The recommended editorial starting point is not one model. It is a stack: strong primary reasoning, cheaper fallbacks, and free or local capacity for low-value tasks.

Deployment

OpenClaw can start on a laptop, in Docker, or on a low-cost cloud instance

Model cost usually matters more than server cost, but deployment choice still shapes setup speed, IM integrations, and how much control you keep.

China-first one-click path

Aliyun, Tencent Cloud, and Volcengine lowered the entry cost dramatically

For many Chinese users, one-click cloud deployment is easier than hand-built VPS setup. For global self-hosting, Docker and a simple VPS remain the cleanest operator path.

from
from ¥9.9 / mo

One-click cloud is easiest. Hetzner-style VPS is still the cleanest global community playbook.

Open deployment hub
Cost Control

The expensive part is usually tokens, not the server

A cheap server with an uncontrolled model setup can still burn hundreds of dollars. Fallback chains and daily budgets are not optional best practices.

baseline

OpenClaw itself is free and open source

Your real bill comes from API tokens, always-on workloads, cron loops, and whether you keep a server running 24/7.

free
Local or free-tier models
$0 API cost

Use Ollama, GLM Flash, ERNIE Speed, or Gemini Flash for experiments, heartbeat jobs, and low-value tasks.

recommended
Hybrid stack
$5 to $20 / mo

Claude Sonnet for difficult work, cheaper fallbacks for daily traffic, and a hard daily budget ceiling.

risk
No fallback + no budget
$100 to $1,000+ / mo

This is how people wake up to frightening bills after an overnight loop or a noisy automation task.

FAQ

Questions people ask before they commit to an OpenClaw setup

These answers reflect the March 2026 security context and the site's unofficial community perspective.

Start with the first blocker, not the whole ecosystem at once

If OpenClaw is not running, go to deployment. If it is running, the next real decisions are usually channels, model stack, or security and cost control.