What is OpenClaw?
OpenClaw is a free, MIT-licensed open-source personal AI agent you self-host. It runs an agentic loop with persistent memory, connects to chat apps like Telegram and WhatsApp, and works with any major LLM.
OpenClaw is an open-source framework for building your own personal AI agent, released under the permissive MIT licence. Its tagline is “the lobster way” 🦞, and the project has grown to 15,000+ GitHub stars and 200+ contributors. Unlike a chatbot you prompt one question at a time, OpenClaw runs continuously: it remembers context across conversations, follows a schedule, and can take actions on your behalf through the messaging apps you already use.
The trade-off is that it is built for tinkerers. You run it on your own infrastructure, define its personality and rules in config files, and own the ongoing maintenance. That control is the point for some people and the obstacle for others.
How it works
OpenClaw wraps a large language model in an agent loop: it reads incoming messages, decides what to do, calls tools or “skills”, writes results back, and stores what it learned for next time. It is LLM-agnostic, so you can point it at Claude, GPT, Gemini, or DeepSeek depending on cost and capability.
Its behaviour is driven by plain-text config files rather than code. SOUL.md defines the agent’s identity and personality, HEARTBEAT.md sets its schedule, AGENTS.md and USER.md capture working context, and separate memory files persist what it learns. See our breakdown of SOUL.md for how that file shapes the agent.
On the channel side it speaks Telegram, WhatsApp, Slack, Discord, Signal, iMessage and more, so the agent meets you where you already chat.
Worked example
The same agent loop can be self-hosted or consumed as a managed service. The difference is who carries the operational weight.
| Dimension | Self-hosted OpenClaw | Liv (OpenClaw as a service) |
|---|---|---|
| Setup | ~15 min (Docker/Node) to a few hours tuning | 14-day trial, no card |
| Infra + LLM cost | Typically under $100/month, you manage it | Pro $79/mo, Max $149/mo |
| Ops | ~1–3 hours/month (tokens, updates, debugging) | Managed for you |
| Security/compliance | Your responsibility | CASA Tier 2, encrypted vaults |
Try this in Liv
Liv is OpenClaw’s agent loop and real integrations wrapped in managed infrastructure, so you get the agent without the ops.
- Start a 14-day free trial at app.liv4all.com, no credit card needed.
- Message Liv on Telegram, the default and required channel.
- Connect Gmail and Calendar via Google OAuth (revocable any time).
- Optionally link WhatsApp (invite-only, needs a dedicated eSIM).
Onboarding is currently early access and batched, so you may join a queue.
Common questions
Is OpenClaw free?
Yes. The software is MIT-licensed and free to use; you pay only for the server it runs on and your LLM API usage. See is OpenClaw free? for the detail.
What can I run it on?
Any OS. Minimum specs are roughly 2 cores, 2 GB RAM and 2 GB storage. System requirements covers this fully.
How do I install it?
Via Docker or Node in about 15 minutes, or a Railway one-click template in about 5. See how to deploy OpenClaw.
Do I have to write code?
No. Configuration is plain-text Markdown files, not programming, though comfort with a terminal helps.
Is Liv the same as OpenClaw?
Liv is an independent managed service built on OpenClaw. It is not affiliated with or endorsed by the OpenClaw project.
What if I don’t want to run a server?
Use a managed option. OpenClaw hosting compares the choices.