ChatGPT Dots vs OpenClaw: An Always-On Agent Without the $100/Month Pro Plan

Short answer: ChatGPT dots and OpenClaw are the same idea — an always-on AI agent you message from your chat apps — with opposite trade-offs. Dots run on OpenAI's cloud, are managed for you, and start at the ChatGPT Pro plan ($100/month). OpenClaw is open source, runs on a machine you control, and — as of this week — can run on the usage included in a $20/month ChatGPT Plus plan. You are choosing between convenience you pay OpenAI for and control you take responsibility for.

This page compares them on the questions that actually decide it: cost, what each one can touch, how safe each is, and whether you can self-host. Facts about dots come from OpenAI's own pages, checked on September 30, 2026; sources are listed at the end. We are an independent guide, not affiliated with OpenAI or the OpenClaw project.

What OpenAI actually announced

OpenAI introduced dots on September 29, 2026. According to the official announcement:

• What a dot is: an always-on agent powered by GPT-6 Astra, with its own cloud computer and browser, connecting to 4,000+ apps through the ChatGPT plugin ecosystem.
• Where you reach it: ChatGPT (web, mobile, desktop), SMS, Slack, and Teams. Pro users can join a limited waitlist for iMessage and RCS.
• Who gets it: rolling out to Pro and Business Premium users in eligible markets; Enterprise users can try a beta once a workspace admin enables it.
• What it costs: your first dot is included in Pro or Business Premium at no extra charge. OpenAI says you will later be able to pay a fixed monthly fee to add more dots or increase their speed and capacity — no price for that has been announced.
• How limits work: conversations with your dot do not count toward your ChatGPT usage limits. Tasks your dot starts or manages in Codex or ChatGPT Work do count, as usual. The first month after launch comes with higher limits.

ChatGPT Pro is currently offered in $100, $200, and $500/month tiers on the ChatGPT pricing page, and the announcement names Pro without singling out a tier, so the cheapest personal route to a dot is $100/month. Some early coverage quoted $200/month; check the pricing page for the tier you are on.

ChatGPT dots vs OpenClaw vs Manus vs Claude

Manus and Claude are in the table because they are the other names people compare against; neither is a like-for-like "always-on agent in your messaging apps" in the same way dots and OpenClaw are. Check their pricing pages directly — both change often and show regional prices.

The cost comparison, done honestly

The new fact that changes this comparison is on OpenAI's help page Using your ChatGPT plan in other apps and sites. It lists OpenClaw among the open-source tools where you can sign in with ChatGPT and, on Plus or Pro, let requests use your ChatGPT plan instead of a separate API key.

So the two realistic paths look like this:

• Dots: ChatGPT Pro at $100/month (or more). One dot included; conversations with it do not use your ChatGPT limits.
• OpenClaw: ChatGPT Plus at $20/month, plus hardware. On your own laptop that is $0 extra but only while it is awake. For always-on, our VPS guide estimates roughly $4–6/month for a small server — that is our estimate, not an OpenAI number.

On paper that is roughly $100 versus $25 a month. It is not a like-for-like swap, and the help page says why:

• OpenClaw's requests count toward the ChatGPT Work and Codex usage included in your plan. An agent that runs all day can use up a Plus plan's weekly allowance much faster than you would by chatting.
• You can — and should — set a weekly usage limit per app under Settings → Usage. That limit is a cap, not an extra pool: it does not add usage to your plan.
• If you run out, the app stops using your plan unless you have separately allowed apps to spend ChatGPT credits. That setting is off by default; leave it off until you know your usage.

If your agent does light, bursty work, the OpenClaw path is dramatically cheaper. If you want it doing hours of real work every day, compare against Pro honestly — the $100 tier includes far more usage, and dot conversations do not draw on it at all.

Is ChatGPT dots safe?

OpenAI describes more built-in guardrails for dots than most self-hosted setups have on day one. From the announcement:

• Isolation: each dot works on its own cloud computer; your computer stays separate unless you choose to connect it. For supported sign-ins, a dot can use saved passwords without exposing them to the model.
• Restricted background work: "proactive research" uses your connected apps through restricted tools that cannot send messages, change app content, or control your browser or computer.
• Rules and approvals: built-in rules decide when a dot may act alone and when it must ask; custom rules let you allow, require approval for, or forbid specific actions. Some sensitive tasks, such as changing passwords, always require you.
• Monitoring: safeguards against malicious instructions are on by default, and a dot can pause or stop work when it detects a risk.
• Training: content from Business, Enterprise, and Edu workspaces is not used to improve models by default; on personal plans you control whether dot conversations and work are used.

OpenAI also states plainly that dots can still make mistakes and that you should review high-impact work. That is the honest summary: dots are designed carefully, but "safe" still depends on what you connect. A dot with access to your email and calendar can act on your email and calendar. The difference from OpenClaw is where the trust sits — with dots, you trust OpenAI's cloud, rules engine, and data policy; with OpenClaw, you trust yourself and code you can read.

Is OpenClaw safer or riskier than dots?

Neither, by default — the risk moves. OpenClaw gives you the controls (allowed-sender lists, channel scoping, a security audit command) but you have to use them, and on a public VPS you also own server hardening. Our OpenClaw safety assessment walks through the real risks and a first-session checklist.

Pick OpenClaw for control and inspectability; pick dots if you would rather a large provider carry the security engineering and you accept their cloud holding your context.

Can you self-host ChatGPT dots?

No. A dot works on a cloud computer that OpenAI provides. You can let a dot connect to your own device, but the agent itself runs in OpenAI's service. If running everything on hardware you control is the requirement, OpenClaw is the option on this page that meets it.

Which should you choose?

• Already on Pro, or on Business Premium? Try your included dot first; it costs you nothing extra.
• On Plus and do not want to pay $100/month? OpenClaw on your ChatGPT plan is the direct ChatGPT dots alternative. Start locally with our install guide, set a weekly app limit, and move to an always-on server only if you need it.
• Need it always on without managing a server? Compare a hosted OpenClaw provider using the five questions in our hosted vs self-hosted guide.
• Handling sensitive data and need to inspect the code? OpenClaw, self-hosted, scoped tightly.

Sources

Checked on September 30, 2026. Prices and availability change; the official pages are authoritative.

• OpenAI — Introducing dots: model, channels, eligible plans, included dot, usage-limit rules, safety measures
• ChatGPT — Pricing: Pro tiers of $100, $200, and $500/month
• OpenAI Help Center — What is ChatGPT Plus?: $20/month
• OpenAI Help Center — Using your ChatGPT plan in other apps and sites: OpenClaw listed; Plus/Pro only; Work and Codex usage; per-app weekly limits
• OpenClaw — documentation and source code

"ChatGPT" and "dots" are OpenAI products; "OpenClaw" is an independent open-source project. This site is an unofficial guide.