Dots is OpenAI’s managed personal agent in ChatGPT. It has a cloud computer and carries context across connected channels. Beans runs a team of bots on your assigned computers with direct provider connections.
You want named bots with direct model connections, group conversations, and encrypted pairing between computers and phones. You are comfortable keeping the assigned computer online.
Cons
You maintain the Runner and keep it online. Model usage costs separately, and the app has no built-in graphical browser.
OpenAI Dots
Pros
You want a managed agent inside ChatGPT, connected apps and background work without maintaining a Runner.
Cons
Access depends on plan and market. You use OpenAI’s service, and multi-dot teams remain a future feature.
Twelve practical differences, based on the mechanisms described in each product’s documentation.
Where work runs
Beans
An assigned Mac, Windows, or Linux Runner that you own. It needs to remain online.
OpenAI Dots
OpenAI-managed cloud computer; optional connection to your devices.
Model connections
Beans
ChatGPT, Grok, Anthropic, DeepSeek, OpenCode, and compatible model endpoints.
OpenAI Dots
OpenAI service powered by GPT-6 Astra.
Subscriptions and APIs
Beans
ChatGPT sign-in; production Grok OAuth disabled. API keys for other providers. Credentials sync encrypted to paired devices.
OpenAI Dots
Eligible ChatGPT plans; not a bring-your-own-provider interface.
Named teammates
Beans
Persistent bot names, instructions, provider selection, memory, and an assigned Runner.
OpenAI Dots
A named primary dot; specialist dots are previewed separately.
Collaboration
Beans
Direct messages, groups of up to six bots, mentions, and handoffs across Runners.
OpenAI Dots
Delegates work to Codex and ChatGPT Work; dot teams are announced as future work.
Files and tools
Beans
Folder-based file work, shell commands, web search, and MCP plugins with permission flows.
OpenAI Dots
Cloud computer and permissioned plugins.
Browser work
Beans
Web search and page reading. The current app does not provide a built-in graphical browser.
OpenAI Dots
Its own browser; optional access to your connected computer.
Persistent context
Beans
Bot notes and chat history on your devices, with encrypted sync through the relay.
OpenAI Dots
Retained context and configurable sharing with ChatGPT Memory.
Scheduled work
Beans
Bot routines run on the assigned computer, which must be online when work is due.
OpenAI Dots
Always-on work and read-only proactive research.
Other devices
Beans
Pair computers and phones. Phones let you chat; desktop Runners execute the tools.
OpenAI Dots
ChatGPT desktop, web and mobile, Slack and Teams.
Data and privacy
Beans
The relay carries ciphertext. Paired devices decrypt it; your selected model receives sent prompts.
OpenAI Dots
Service-side encryption at rest and in transit; plan and training settings apply.
What you pay
Beans
Free app. Provider subscriptions/API usage and your computer’s operating costs are separate.
OpenAI Dots
First dot included in eligible paid plans, with work allowances and delegated-task limits.
What these differences mean
Execution changes your everyday workflow.
A Beans bot has one assigned Runner. It works in that computer’s folders with that computer’s user permissions. Choosing a local workspace or a managed cloud machine changes which files are available, what stays online, and who maintains the environment.
Pick the provider without replacing the teammate.
Beans stores a bot’s role separately from its model choice. Provider credentials belong to your account and sync encrypted to paired devices. A subscription login and an API key are different billing arrangements; having one does not automatically include the other.
Collaboration is more than parallel windows.
Beans has one-to-one chats and group chats with up to six bots. Mentions direct a turn to a bot, and handoffs pass work between teammates. Different Runners can contribute to the same chat, so the conversation can span your computers.
Tools use the permissions you give them.
Bots can work with files, execute commands, and use connected MCP tools. Review the app’s permission flows and the tools you install. Running a command with your user’s permissions is different from isolating it in an operating-system sandbox.
Separate device privacy from model privacy.
Beans encrypts account data before relay sync. The relay cannot decrypt the chat blobs. Inference still sends content to the model provider you choose. Review both the sync mechanism and that provider’s data policy when deciding what to share.
Plan for work while you are away.
Your phone lets you talk to bots on paired computers. It does not replace their Runner. Routines and new turns need the assigned computer online. For always-on work, choose a computer or server you can keep running and account for its maintenance.
Separate the app from the AI bill.
Beans is free. That does not make model usage free. Each bot uses its connected provider, and parallel work can consume more allowance. A Runner needs to be online; using your own server also has operating costs.
Beans app
$0
AI provider
Your subscription or API usage
Running computer
Your hardware, power, and connectivity
OpenAI Dots First dot included in eligible paid plans, with work allowances and delegated-task limits.
Which fits your work?
Choose Beans when…
You want named bots with direct model connections, group conversations, and encrypted pairing between computers and phones. You are comfortable keeping the assigned computer online.
Does encrypted sync mean the model cannot see my prompt?
No. Device sync and model inference are separate. Beans encrypts data between paired devices, but your selected AI provider processes the content you send it to generate a reply.
Can I keep my existing AI subscription?
Beans offers ChatGPT subscription sign-in. Production Grok OAuth is disabled pending a verified Beans contract. Other providers use API keys; check their access, plan and usage requirements.
Does my phone run the bots?
No. A paired phone is a chat device. Each bot runs on an assigned Mac, Windows, or Linux Runner. Work waits if that Runner is offline.
Are these performance rankings?
No. These pages compare documented mechanisms and workflows. They do not claim speed, quality, token savings, or reliability measurements.
Sources, scope, and updates
Checked October 7, 2026. Product capabilities, availability, plans, and limits can change. Consult the linked official documentation before purchasing or migrating.
Explore the actual desktop interface in the homepage demo, then check the published downloads for your device. The demo uses sample data and simulated replies, not a connected AI provider.