ChatGPT Dots vs OpenClaw

The choice in ChatGPT dots vs OpenClaw comes down to whether you want a managed agent inside OpenAI's subscription or a self-hosted assistant running on your own hardware. Each dot runs on its own cloud computer, follows Custom Rules for when to ask your approval, and connects to over 4,000 apps through OpenAI's plugin ecosystem. OpenClaw is open-source software under the MIT license that runs on your own computer or a server you host, with the model provider you choose, including local models.

A dot is included at no extra cost in ChatGPT Pro starting at $100 per month or ChatGPT Business Premium at $125 per user monthly, with no server maintenance required. OpenClaw software is free, but you pay your model provider for API tokens, plus server hosting if you do not run it on your own machine.

OpenAI launched dots on September 29, 2026, and access is rolling out gradually to eligible accounts. Because dots are brand new, real-world operational benchmarks are not in yet. Everything said here about dots is OpenAI's own published description, set against OpenClaw's own site, README and docs.

ChatGPT dots vs. OpenClaw: Side-by-Side

Are you one of these vendors? Update your listing

Which One to Pick

Pick ChatGPT dots if you pay for Pro or a Business Premium seat and want a managed agent with approval rules and no server to run; pick OpenClaw if you want to self-host, choose your own model, and avoid OpenAI's plan and region limits.

The deciding question is who runs the machine. A dot works on its own cloud computer that OpenAI operates, while OpenClaw runs on hardware you own or rent, with the model provider you pick.

OpenClaw, maintained by the OpenClaw Foundation under the MIT license, is an open-source assistant you install on your own computer or a server you host. It uses a hosted or local model you choose and answers you in chat apps such as WhatsApp, Signal and Discord.

The older dot icons in the ChatGPT interface are unrelated to these agents; see what the ChatGPT dot means.

• Pick ChatGPT dots for no server maintenance and the ability to start Codex and ChatGPT Work tasks from your dot.
• Pick OpenClaw for self-hosting, credentials kept on your own host, and access through chat apps like WhatsApp or Signal.
• OpenAI names GPT-6 Astra as the model behind dots and has published no model choice, while OpenClaw treats model providers as swappable plugins.
• Dots include approval controls through Custom Rules, whereas OpenClaw relies on manual configuration and local pairing codes.

Operating Costs and Plan Pricing

ChatGPT dots require an active premium subscription starting at $100 monthly, while OpenClaw distributes free software that incurs variable infrastructure and model token fees.

Your first dot is included in ChatGPT Pro or ChatGPT Business Premium at no extra cost. Pro tiers include Pro 100 at $100 per month, Pro 200 at $200 per month, and Pro 500 at $500 per month, all billed monthly with no annual discounts. ChatGPT Business provides Standard seats at $25 per user monthly ($20 annual) and Premium seats at $125 per user monthly ($100 annual), requiring a minimum of two seats. Check the OpenAI pricing page, the OpenAI Pro tiers article, and the ChatGPT Business billing article to verify current figures, as subscription terms change over time.

OpenAI published that eligible plans receive an allowance for deeper work, with extended limits for the first month after launch, and dot usage will not count toward plan allowances for the next month. However, VentureBeat reported that OpenAI has not disclosed numeric allowances, additional dot pricing, or post-first-month terms.

OpenClaw publishes its software for free at openclaw.ai with no subscription and no hosted commercial tier. Users manage their own expenses, which combine token fees for model endpoints and optional hosting fees. In our pricing analysis at OpenClaw pricing, running costs generally range from roughly $0 on local models, $10 to $30 monthly for light personal assistant use, $20 to $50 monthly for a lean business deployment, and $300 to $600 monthly for continuous operation on high-end commercial model endpoints.

• ChatGPT Pro 100: $100 per month includes your primary dot and reasoning powered by GPT-6 Astra.
• ChatGPT Business Premium: $125 per user monthly (or $100 per user monthly billed annually) includes your first dot in all supported ChatGPT regions.
• OpenClaw software: $0 license fee, but requires external model token budgets and compute resources.
• OpenClaw's running cost depends on the model provider you pick; our planning range for local models is roughly $0.

→ Compare per-token pricing for every AI model

Setup Requirements and Server Infrastructure

ChatGPT dots deploy through a guided onboarding flow in the ChatGPT desktop application, whereas OpenClaw installs from the command line or npm onto a computer or server you maintain.

To set up a dot, you open the ChatGPT desktop application on macOS or Windows, or access ChatGPT on desktop web, name your primary dot, choose an avatar, and connect your everyday work tools. The agent operates entirely on its own cloud computer with its own browser, continuing work 24/7 without keeping your personal computer running. Local computer access is optional, starts turned off, and must be explicitly approved in the desktop app if you want your dot to access local files or use local desktop software.

OpenClaw installation takes place through a shell script or package manager, as documented on the OpenClaw GitHub repository. The architecture runs a local Gateway process that binds to loopback, coordinating communication between chat channels and language models. Running it on a server instead of your own machine is supported, and it adds hosting cost and a larger security surface.

• ChatGPT dots require zero infrastructure provisioning or server maintenance.
• Dots cloud computers run independently, meaning your laptop can shut down while tasks complete.
• OpenClaw runs on your own computer, or on a server you pay to host.
• OpenClaw's installer sets up Node.js and the rest; credentials then live in .env files and local config on the host.
• OpenClaw's docs say to treat third-party skills as untrusted code and read them before enabling.

Supported Models and Provider Flexibility

OpenAI says dots are powered by GPT-6 Astra and has not published a way to pick another model, whereas OpenClaw treats hosted and local model providers as swappable plugins.

OpenAI built dots around GPT-6 Astra, giving the agent native hooks into Codex cloud environments for software engineering and ChatGPT Work for document drafting and multi-step research. OpenAI has not published any setting for choosing a different model for your dot.

OpenClaw decouples agent orchestration from the underlying language model. OpenClaw's README names Claude, Codex and local models as providers you can swap in. On using ChatGPT to power OpenClaw, Codex, OpenAI's coding agent, is the OpenAI option the README names; it does not list ChatGPT or dots.

• ChatGPT dots are powered by GPT-6 Astra; OpenAI has published no model choice.
• Dots integrate directly with OpenAI Codex and ChatGPT Work for code and artifact generation.
• OpenClaw supports swappable model providers, allowing users to alternate between Claude, Codex, and local models.
• OpenClaw can run on local models, so prompts need not go to a hosted model provider.
• OpenClaw's README names Codex as its OpenAI option and does not list ChatGPT itself.

Execution Control and Security Posture

OpenAI isolates ChatGPT dots inside monitored cloud environments with structured approval gates, while OpenClaw runs tools directly on the host machine and stores configuration credentials in local text files.

Each dot works on an isolated cloud computer, keeping your local machine separate unless you grant access through the desktop app. OpenAI implements Auto-review, which evaluates actions against your instructions, Custom Rules, and safety requirements. Custom Rules provide four distinct behaviors: 'Take action without asking', 'Take action if pre-approved', 'Ask before taking action', and 'Hand off to you'. Critical actions like changing passwords always stay with the human user, and saved passwords can be used on supported sites without exposing them to the model. Proactive research is limited to read-only tools that cannot alter app contents or send messages. OpenAI notes that dots can still make mistakes, advising users to review consequential work.

OpenClaw's defaults are set out in its gateway security docs, and whether OpenClaw is safe for business walks through the risks. On most chat channels, an unknown direct-message sender gets a pairing code instead of having the message processed, and group access is allowlisted. OpenClaw's site says it completed a security audit through OpenAI's Patch the Planet initiative. However, tools execute directly on the host machine unless sandboxing is configured, and API credentials reside in local .env files. Security researchers found hundreds of malicious skills on ClawHub in 2026, as Unit 42 reported.

At Layer3Labs, we build and run AI systems inside other people's businesses, and every rollout we have done starts with a list of which accounts and keys the agent can reach. On OpenClaw that list starts with the .env files on the host, because tools run on that host unless you configure sandboxing.

• Each dot works on its own cloud computer, with Auto-review and Custom Rules deciding what needs your approval.
• Dots allow users to set granular approval gates: take action, pre-approve, ask first, or hand off.
• OpenClaw tools run directly on the host operating system unless sandboxing is explicitly enabled.
• OpenClaw credentials sit in local text configuration files rather than managed identity vaults.
• Public OpenClaw skill registries have suffered supply chain attacks, requiring manual code review before skill installation.

Supported Chat Channels and Connected Apps

OpenClaw connects to over 20 consumer and workplace messaging networks, while ChatGPT dots focus on ChatGPT interfaces alongside enterprise connectors for Slack and Microsoft Teams.

ChatGPT dots meet users inside existing OpenAI touchpoints, including desktop applications, desktop web, and mobile apps. OpenAI provides enterprise integrations for Slack and Microsoft Teams, allowing teams to message dots directly. You can hop on a voice call with your dot, though it cannot initiate calls to you at launch. SMS texting is in a limited beta for United States Pro users. Dots connect to over 4,000 applications through OpenAI's plugin ecosystem, with permissions managed inside the ChatGPT profile under Customize.

OpenClaw focuses heavily on universal communication channels. According to the project's documentation, OpenClaw supports 29 channels including WhatsApp, Telegram, Discord, Signal, iMessage, Slack, and Microsoft Teams. Integrations come from community skills published on ClawHub or skills you write yourself. The official OpenClaw skills documentation instructs users to treat third-party skills as untrusted code and use openclaw skills verify to review scan reports before activation.

• ChatGPT dots communicate via ChatGPT apps, Slack, Microsoft Teams, and limited United States texting.
• Dots connect to over 4,000 business applications using OpenAI's established plugin catalog.
• OpenClaw connects natively to WhatsApp, Telegram, Signal, Discord, and iMessage.
• OpenClaw lets you write your own skills, and its docs say to read third-party skills before enabling them.
• You can start a voice call with your dot; it cannot call you at launch.

Regional Availability and Plan Eligibility

OpenAI limits where Pro users can get a dot, while OpenClaw is free software you install yourself, with no subscription or hosted tier to qualify for.

OpenAI began rolling out dots on September 29, 2026. For individual subscribers on Pro plans, access is available in markets excluding the EEA, Switzerland and the UK. For business workspaces on ChatGPT Business Premium, access is available across all supported ChatGPT regions. Enterprise workspaces can access the beta when an administrator enables the feature. OpenAI does not list a dot in the Free, Go or Plus plans.

OpenClaw is open-source software distributed on GitHub with no subscription and no hosted tier. It runs on macOS, Linux and Windows with whichever hosted or local model you connect.

• ChatGPT Pro dots are unavailable to subscribers located in the EEA, Switzerland, and the United Kingdom.
• ChatGPT Business Premium provides dot access across all supported ChatGPT international regions.
• Dots are rolling out gradually, meaning access may take several days to appear on eligible accounts.
• OpenClaw has no subscription or hosted tier, so there is no plan to qualify for.
• OpenClaw requires access to third-party model APIs or local hardware capable of executing inference.

Workflows Best Suited for ChatGPT Dots

ChatGPT dots fit people who want a managed agent and already work in ChatGPT, Slack or Teams.

The main draw of dots is that OpenAI runs the computer. If you already subscribe to ChatGPT Pro or Business Premium, your primary dot provides continuous cloud execution without server administration. Knowledge workers who spend their day in Slack, Microsoft Teams, and ChatGPT can delegate research, scheduling, and project monitoring without managing infrastructure.

Dots also suit teams that want approval steps written down. Custom Rules let you allow, require approval for, or block specific actions, and Auto-review checks account-affecting actions against those rules, though OpenAI says a dot can still make mistakes, including when following your rules.

• Teams already paying for ChatGPT Pro or ChatGPT Business Premium looking for an out-of-the-box agent.
• Workers who want background research and task management without keeping a home computer powered on.
• Organizations that demand automated safety checks and explicit approval workflows before actions execute.
• Businesses operating primarily in Slack, Microsoft Teams, and OpenAI's workspace tools.

Environments Best Suited for OpenClaw

OpenClaw fits developers and technical operators who want to host the agent themselves and pick the model.

Because OpenClaw runs on your hardware or a server you host, credentials and config stay on that machine, and with a local model your prompts stay there too. Developers can inspect every line of the MIT-licensed codebase and modify agent behavior directly.

OpenClaw also suits people who want an agent in WhatsApp, Signal or Telegram, or who would rather pay a model provider directly than a monthly ChatGPT plan.

• Engineers and system administrators comfortable managing command-line installations and server daemons.
• Users in the European Economic Area, Switzerland, or the United Kingdom seeking an agent on individual budgets.
• Operators who require strict data privacy and local model execution without cloud vendor data processing.
• Workflows that rely on personal messaging apps like Signal, WhatsApp, Telegram, or Discord for daily tasks.

The Verdict

Choose ChatGPT dots if you already subscribe to ChatGPT Pro or Business Premium, operate in an eligible region, and want an autonomous cloud coworker that runs 24/7 without server management. You get a managed cloud computer, Custom Rules for approvals, and connections to over 4,000 apps, with the caveat that real-world reports on dots are not in yet.

Choose OpenClaw if you are a developer or technical operator who demands complete ownership over your software, no OpenAI plan or region limits, multi-channel messaging access via WhatsApp or Signal, and the freedom to route tasks across swappable model providers or local hardware.

For individuals on Free or Plus plans who do not want to spend $100 per month, and who lack the technical appetite to maintain a private server, the right path is staying with standard conversational chat and manual automation tools. If OpenAI expands dots to lower-priced tiers or if OpenClaw releases a managed cloud version with automated skill sandboxing, this balance will shift.