Comparison
Belowdecks and OpenClaw compared point by point
OpenClaw is an open source AI assistant, for a person or a team, that you install and maintain yourself. Belowdecks installs AI agents in your business, and our team sets them up and oversees them. This page compares the two for a small business.
Two ways to hand work to AI
OpenClaw is an open source AI assistant, under the MIT license, run by the OpenClaw Foundation, a nonprofit. You install it on a computer or a server, then it answers in your messaging apps, such as WhatsApp, Telegram, Slack or Microsoft Teams. It can also work on a schedule, read a mailbox and keep what it learns in memory files. The project had more than 390,000 stars on GitHub as of October 5, 2026.
You install, configure and update it yourself. The installer sets up Node.js and launches a setup wizard, but configuring each messaging app, the access rights and the sandbox requires being comfortable with a terminal. Several versions come out every week. OpenClaw’s documentation describes a gateway as a single trust domain: everyone who writes to an agent shares the access that agent was given.
Belowdecks installs AI agents in your business that take on part of the work, and our team sets them up and oversees them every month. With you, it decides what starts each agent and which actions must wait for approval by someone on your team. OpenClaw therefore suits a person or a technical team that wants to host and configure its own assistant, and Belowdecks suits a business that hands work to agents and wants a team to take care of them.
OpenClaw and Belowdecks, line by line
The facts about OpenClaw come from its GitHub repository, its documentation and the security advisories published about it, checked on October 6, 2026. The project ships several versions a week, so these facts can change quickly.
What you get
- OpenClaw
- A free, open source assistant for a person or a team
- Belowdecks
- Agents installed and monitored by a team
Who builds and maintains it
- OpenClaw
- You: installation, settings, updates
- Belowdecks
- The Belowdecks team, from diagnostic to follow-up
What starts the work
- OpenClaw
- Messaging apps, schedule, Gmail, IMAP, webhook
- Belowdecks
- An email, a schedule or a signal from your software
Your go-ahead before an important action
- OpenClaw
- Commands: off by default; the rest in code
- Belowdecks
- Set up at installation to wait for your approval
Use by several people
- OpenClaw
- Named roles; one gateway per group of trusted people
- Belowdecks
- Access by group, each with its own permissions
Assistant for your customers
- OpenClaw
- No widget documented for a public site
- Belowdecks
- Assistant on your site; the customer confirms their request
Where data is hosted
- OpenClaw
- On your machine; text sent to the model
- Belowdecks
- Your Belowdecks workspace; region agreed at the diagnostic
AI bill
- OpenClaw
- Your key, your subscription or a local model
- Belowdecks
- Billed in your name by the AI provider
Isolating the agents
- OpenClaw
- Optional sandbox; skills to vet
- Belowdecks
- Each agent limited to its folder, checked by a test
Monitoring the work
- OpenClaw
- Usage and cost tracking in the interface
- Belowdecks
- A result for each task, a monthly report
Help and support
- OpenClaw
- Online documentation
- Belowdecks
- The Belowdecks team
Pricing structure
- OpenClaw
- Free software; machine and tokens at your cost
- Belowdecks
- From $449 a month, setup quoted separately
Choosing
The situation where each one fits best
Both options meet real needs. The right choice depends mostly on your team’s technical skills and the intended use.
OpenClaw fits better if
You want your own assistant in your messaging apps, on your hardware, with the model of your choice. You or someone on your team knows how to isolate the machine that runs it, configure its access and apply frequent updates.
Belowdecks fits better if
You want to hand work to agents, with access by group for your team and an assistant for your customers on your website. Important actions wait for your go-ahead, and our team installs, updates and monitors the agents.
Common questions about Belowdecks and OpenClaw
OpenClaw is free. What does Belowdecks charge for?
The OpenClaw software is free, and the project offers neither a paid plan nor a hosted service. The real costs are the machine that runs it, the model’s tokens and the time of the person who installs it, configures it and applies the updates. The documentation notes that the agent’s heartbeat, a full turn every 30 minutes by default with an API key, every hour with a subscription login, uses tokens each time.
With Belowdecks, plans start at $449 a month with Crew, which includes 15 million tokens, and the setup is quoted separately. It includes the agents our team designs for you, and the number of agents does not change the price. The AI provider bills you directly for what the model consumes. The Pricing page details the three plans.
What do the security advisories say about OpenClaw?
In February 2026, Koi Security researchers found 341 malicious skills among 2,857 analyzed in ClawHub, OpenClaw’s community registry. The registry has since shown a scan status for each skill. The same month, Microsoft wrote that OpenClaw is not appropriate for a standard personal or enterprise workstation, and recommended a virtual machine or a separate machine, with accounts and tokens created solely for the agent.
In May 2026, the Cyber Security Agency of Singapore recommended not installing it on a device that holds sensitive data and requiring human approval for high-stakes or irreversible actions. The project publishes its security advisories and their fixes on GitHub.
Can OpenClaw ask for approval before acting?
Yes, for the commands the agent runs on the machine. These prompts are off by default, and you can require them for every command or for those outside your allowlist. You answer in the interface, the command line or the chat. For other actions, such as sending an email or publishing, you have to write a plugin that asks for approval.
In Belowdecks, our team sets up with you, at installation, the actions that must wait for approval by someone on your team.
Can we put OpenClaw in front of our customers?
The documentation provides no chat widget for a public website, and the included WebChat is the interface of the person who runs OpenClaw. Direct messages from strangers can be opened through an explicit setting. The documentation notes, however, that everyone who writes to an agent shares its access, and recommends one gateway per group of people who don’t trust each other.
On your website, the Belowdecks assistant shows each customer only their own conversation, and that customer confirms their own request.
Sources
The information about OpenClaw comes from sources checked on October 6, 2026: the GitHub repository, the documentation on team use, approvals, WebChat and the heartbeat, an article in The Hacker News dated February 2, 2026, a Microsoft blog post dated February 19, 2026 and an advisory from the Cyber Security Agency of Singapore dated May 28, 2026. The project changes every week, so check this information before you decide.
OpenClaw is a project of the OpenClaw Foundation. Belowdecks is not affiliated with this project.
First call
Find out whether some of your work suits agents
A free 20-minute call to describe the work that takes up your time. We will tell you whether Belowdecks fits, or whether an assistant like OpenClaw is enough.