What OpenClaw is, and how to run it 24/7
OpenClaw is an open-source AI agent: it reads your instructions, calls a model, uses tools and skills, and talks to people through its own gateway and chat channel plugins. Version 2.0 shipped on August 30, 2026. Getting an OpenClaw agent working takes an afternoon. Keeping it working around the clock is a standing commitment: a machine that never sleeps, the updates, the secrets, the restart after a reboot, the spend nobody capped. This page covers what OpenClaw is and the three places it can live.
Direct answer
What is OpenClaw, and how do you run it 24/7?
OpenClaw is an open-source AI agent with its own gateway, chat channel plugins, and skills; its current major release, 2.0, shipped on August 30, 2026. To run it 24/7 it needs a machine that never sleeps. A laptop is fine for experiments but stops when the lid closes. A VPS keeps it online if you operate the server yourself. Managed hosting, such as Qoren, operates the environment, updates, and monitoring for you.
OpenClaw deployment options
| Option | Operations burden | Best fit |
|---|---|---|
| Local machine | You keep the machine awake, patched, and backed up. | Free experiments and local-only workflows. |
| VPS | You manage Linux, Docker, firewalling, updates, logs, and recovery. | Developers who want full host control. |
| Managed hosting | The provider operates the cloud environment and runtime updates. | Always-on OpenClaw agents without server maintenance. |
What OpenClaw is
OpenClaw is an open-source agent runtime: the software that reads an agent's instructions, calls the model, uses tools, and keeps its memory. A gateway process sits in front of the agent, and chat channel plugins let people reach it from the apps they already use. Skills and configuration are OpenClaw's own, which is why teams that already run OpenClaw agents usually want to keep them on OpenClaw rather than rebuild them on another runtime.
- Open source, with its documentation at docs.openclaw.ai.
- A gateway process that has to stay up for the agent to answer.
- Chat channel plugins, so people reach the agent where they already talk.
- Its own skills and configuration, which move with the agent.
Sources: OpenClaw documentation
What changed in OpenClaw 2.0
OpenClaw 2.0 shipped on August 30, 2026. It is the largest release in the project's history, with 933 contributors and more than 16,000 pull requests. For anyone running an agent around the clock, five changes matter most. The OpenClaw 2.0 guide linked below covers each one and the safe upgrade path.
- A rebuilt browser app that opens straight into a conversation with your agent, with setup and live monitoring in the same place.
- Guided onboarding that finds the ChatGPT, Claude, and Codex subscriptions, API keys, and local models you already have, and verifies the model works before saving it.
- Shared cloud sessions: bring another person into a live agent session, or hand a task over with its context intact.
- Sessions and transcripts moved to SQLite. It is sturdier for 24/7 use, and it is a one-way door for downgrades, so take a verified backup before you upgrade.
- A reliability pass across Gateway startup, credentials, plugins, mobile pairing, and messaging channels.
Sources: OpenClaw 2.0 release notes
How to run OpenClaw 24/7
An always-on agent needs two things: a machine that never sleeps, and someone doing the upkeep, meaning updates, secrets, restarts, logs, and a spend limit. There are three places OpenClaw can live, and they differ mainly in who does that upkeep.
- Your laptop: free and fast for experiments, but the agent stops when the lid closes, the wifi drops, or the machine reboots.
- Your own VPS: stays online, and you own Linux, the runtime install, firewalling, updates, backups, and noticing when the agent dies. A small machine is $12-18 a month, plus your hours.
- Managed hosting: the environment, updates, and monitoring are handled for you. On Qoren it starts at $39 a month, and the managed OpenClaw hosting page prices every line against self-hosting.
What to decide before the agent runs unattended
Wherever it runs, an unattended agent is easier to trust when the operating contract is clear before it goes live.
- Model provider, and whether the agent uses a managed key or your own.
- Secrets the agent needs, such as app tokens, database credentials, or webhooks.
- Persistent storage requirements and what the agent is allowed to write.
- Schedules, triggers, spend limits, and escalation rules.
Where Qoren fits
Qoren is a managed layer around OpenClaw, not a replacement for it. It provisions an isolated cloud environment, keeps the agent online for scheduled and triggered work, keeps secrets off your laptop, and shows usage and status from one dashboard. Managed environments run the tested current runtime by default, so an upgrade like 2.0 is not yours to schedule. If you already run OpenClaw yourself, its skills and configuration keep working the way OpenClaw users expect.
When not to use Qoren
Qoren is not the best answer for every OpenClaw user. If your priority is a free local experiment, offline use, or full control over every host-level setting, local self-hosting or your own VPS may be the better fit.
Related guides
Managed OpenClaw hosting
Managed OpenClaw hosting: what is included, and what it really costs
Managed OpenClaw hosting from $39 a month: what is included, and a line-by-line cost comparison with a VPS, model tokens, and the hours self-hosting takes.
Read guideOpenClaw 2.0 release
OpenClaw 2.0: what shipped, and what it means for running agents 24/7
OpenClaw 2.0 shipped August 30. What changed, what it means for agents running 24/7, and how to upgrade without breaking your stack.
Read guideRuntime decision guide
OpenClaw vs Hermes: which agent runtime should you deploy?
A neutral guide to choosing between OpenClaw and Hermes: how the two agent runtimes differ, and which one to deploy for managed, always-on agents.
Read guideOpenClaw security checklist
How to lock down a self-hosted OpenClaw gateway
A practical checklist for securing a self-hosted OpenClaw gateway: bind to loopback or Tailscale, require auth, handle keys safely, limit the blast radius.
Read guideFrequently asked questions
Is OpenClaw free?
The software is open source, so OpenClaw itself costs nothing. Running it does: the machine it lives on, the model usage it generates, and the time spent keeping it online.
What is new in OpenClaw 2.0?
A rebuilt browser app, guided onboarding that reuses the subscriptions and models you already have, shared cloud sessions, sessions stored in SQLite, and a broad reliability pass. The SQLite move blocks downgrades, so back up before upgrading.
Can I run OpenClaw without a VPS?
Yes. Managed hosting runs OpenClaw in a cloud environment someone else operates, so the agent stays online without a server of your own. It is not designed for users who specifically need local-only or offline self-hosting.
Does Qoren replace OpenClaw?
No. Qoren is a managed hosting and operations layer for running OpenClaw agents online. OpenClaw remains the agent runtime you are deploying.
Can I bring my own model key?
Yes. On Qoren, bring-your-own-key works on every plan: connect your own OpenRouter key and OpenRouter bills you, with no managed usage cap. Or use the managed model key, which draws on your plan's credits and stops at a hard cap.
Run OpenClaw, Hermes or Codex agents 24/7, without the homework.
Qoren runs the machine, the secrets, the updates and the logs, with a hard cap on credit spend. Start from a template and have an agent working today.