Techsona Labs

Grok Bot vs OpenAI dots: AI Coworkers and Alternatives (2026)

Follow a recurring project brief through memory, connected apps, review and cost. See how Grok Bot and dots differ, and where Claude, Google, AWS and open-source alternatives fit.

In this article
  1. Two Ways to Organise Ongoing Work
  2. What Carries Forward to Next Week?
  3. Where the Work Actually Happens
  4. From a Draft to an Action
  5. When the Alternatives Make Sense
  6. When the Job Becomes a Code Change
  7. The Cost of Keeping Work Moving
  8. Where I Would Start
  9. Questions About Access and Capabilities
  10. Sources and Comparison Method

A weekly project brief sounds like a small job until someone has to keep doing it. The notes are in one place, the deadlines in another, and a decision made on Tuesday changes what should be said on Friday. Producing a fluent summary is only one part of the work. Keeping the responsibility moving is the harder part.

That is the useful starting point for Grok Bot vs OpenAI dots. Both go beyond answering a question in the current conversation. They offer ways to carry work across tasks, connected apps and time. The important difference is how that responsibility is organised, and how much of the surrounding setup fits the way you already work.

Imagine giving each product the same weekly brief. Follow what it needs to remember, where it finds the information, how it runs again and when a person reviews the result. That gives the comparison a practical shape before we bring in the other brands.

Grok Bot vs dots: two ways to organise ongoing work

The first difference appears before the assistant opens a document. Are you handing a job to one of several specialised teammates, or asking a single coordinating agent to keep track of several responsibilities? That choice shapes how you explain the work and where you return to correct it.

Grok Bot starts with a team of Bots

Grok Bot is presented as an AI teammate working across apps, websites and tools. Individual Bots can have their own roles and memories. In the weekly-brief example, one Bot could own project reporting while another handles a different responsibility.1, 2

That separation is useful when it makes the work easier to explain and find again. It also has an important limit: Bots on an account share a persistent cloud computer. Distinct Bot roles do not create independent computers, browser sessions or credential boundaries.4 Think about the shared environment as well as the individual conversations.

Dots starts with a coordinating relationship

A dot is a personal agent in ChatGPT for responsibilities that continue between conversations. One dot can track multiple tasks, use background agents and create visible threads for separate work.7, 8 For the same weekly brief, the emphasis is on keeping the project responsibility with the coordinating agent while the individual tasks move ahead.

I would evaluate that continuity by changing something after the first assignment: a deadline, an approved source or the person who must review the brief. The useful result is more than a remembered sentence. It is the affected work being brought up to date, with any unresolved decision returned to you.

Scroll the table sideways on smaller screens.

The documented differences behind the working relationship
Part of the jobGrok BotOpenAI dots
Organising responsibilityBots with distinct roles and memories.2A dot coordinating multiple tasks and background agents.8
ComputerA persistent computer shared by the account's Bots.4A cloud computer and explicit connections to supported local computers.9
Reusable workSkills and routines with schedules or supported events.3Saved notes, schedules and supported event monitoring.8
Development handoffDocumented delegation to Cursor Cloud Agents.2Creating or continuing supported Work and Codex tasks.8

What should carry forward to next week?

The first brief is the easy place to start. The more revealing moment comes when the project changes and the assignment runs again. Does the assistant use the current deadline? Does it follow the agreed format? Does it know which question still needs an answer?

It helps to separate memory from a procedure. “Keep the brief short” is a preference. Finding the approved notes, identifying overdue decisions and producing the brief are steps in a job. Grok Bot documents reusable skills and routines; its guide recommends getting a one-time task working before automating the procedure.3

Dots distinguishes conversation context, relevant ChatGPT memory and its own saved notes. Its ongoing work can use saved schedules and supported event monitoring.8 In either case, explain what should happen on Friday and how you want to hear about a missing document or a decision that is still pending.

This also answers a common misunderstanding about connected apps. Giving an agent access to a folder does not by itself assign a monitoring job. Confirm the routine or task, its trigger and the result it is expected to return. Then make a correction and see whether the next run follows it.

Where does the work actually happen?

A useful brief depends on the right information, which makes connections more important than the length of an app catalogue. Your assignment might require reading one project folder, checking a calendar and preparing a document in a particular account. Each of those is an action to verify.

Grok Bot documents connected apps, custom MCP tools and browser work. Dots uses supported plugins and connected accounts, alongside its browser and computer capabilities.1, 9 Start with the exact read or write you need. A service appearing in the interface is the beginning of that check, rather than proof that every action in that service is supported.

Cloud execution lets work continue while your laptop is off. Local resources add another dependency. For dots, the documentation specifies that a connected computer must be online with the ChatGPT app open for local work.9 A brief that needs a local file or application therefore has a different operating requirement from one using cloud-accessible sources.

Grok Bot's shared computer deserves a similar practical check. Files, browser sessions and command-line credentials can be shared across the account's Bots.4 Give each Bot a clear role, while keeping the account's shared access in mind.

How should a draft become an action?

There is a meaningful step between preparing Friday's brief and sending it to the team. A useful first assignment makes that step explicit: prepare the draft from these sources, identify unresolved points and return it for review before sharing.

Grok Bot documents Auto Review and custom approval rules. Dots documents action review, app permissions and optional custom rules.5, 10 The behaviour depends on the action and the permissions in force. Look at the actual proposed message or changed file, and decide whether the review point fits the responsibility you are delegating.

Ending the assignment matters too. OpenAI distinguishes pausing a dot, stopping delegated work and cancelling a schedule.10 If a project closes, check those separately so that a conversation appearing quiet is not your only evidence that recurring work has ended.

When do the alternatives make sense?

Once the brief is concrete, the other brands become easier to assess. You are looking for the product that fits the sources, deliverable and ongoing responsibility. There is little value in adding an alternative simply because it also calls itself an agent.

Claude Cowork: start with the files and deliverable

Claude Cowork is the relevant Anthropic comparison when the job is a document, analysis or recurring research brief built from files, connectors and skills. Current documentation covers web, desktop and mobile. New sessions on Pro, Max and Ultra run in the cloud, while local-folder access still depends on the desktop connection.14

Scheduled Cowork work can run remotely.15 I would include it when the weekly brief is the main deliverable and your required files are available through that workflow. Compare how easily you supply context, correct the draft and find the finished document again.

Google: tasks in Gemini, formerly Spark

Google's current help page says Spark was renamed to Task. Tasks in Gemini support ongoing work with connected sources and skills. Their schedules include time-based runs, Gmail monitors and topic monitors, although Google says monitoring is unsuitable for time-critical alerts.16, 17

For a Google-centred project, that is a reason to investigate the exact actions the brief requires. Check access first: the cited guide currently describes eligible personal accounts with Google AI Pro or Ultra, plus age, activity-setting and regional conditions. A work or school account is a different route. Google Workspace Studio separately addresses creating and sharing Workspace automations.16, 24

Amazon Quick: business information across connected systems

Amazon Quick is the closer hosted AWS comparison for this business-work assignment. AWS documents connected information, research, deliverables and workflows.25 It becomes relevant when the brief depends on business systems and the question is how well the service can bring that information into a usable result.

The buying route matters. Quick's current pricing separates individual or team routes from organisation plans with an account-level infrastructure fee.26 Match the proposed connections and plan to the actual team before comparing its price with a personal assistant subscription.

Hermes, OpenClaw and Kiro Crew: choose the platform you will operate

The open-source options introduce a different responsibility: somebody must maintain the environment around the assistant. Hermes combines memory and reusable skills with configurable execution backends. Its scheduled agent jobs require the documented gateway to be running.20, 21 That can suit a procedure you want to inspect and adapt, provided the operating work is worthwhile.

OpenClaw provides a self-hosted gateway connecting channels and agents, with persistent scheduling and delivery to configured destinations.18, 19 If the team wants the brief delivered into a particular conversation, that channel-and-runtime setup is part of what you are evaluating.

Kiro Crew is another persistent, open-source agent running locally or remotely on hardware you operate, with memory, recurring work and messaging channels.27 It belongs beside Hermes and OpenClaw in this platform discussion. Model access, hosting, updates and recovery remain part of the cost for all three.

What if the job becomes a code change?

A project brief may uncover a bug that needs fixing. That gives a coordinating assistant a reason to hand work to a development agent, and it changes what a fair comparison looks like. Grok Bot documents handoff to Cursor Cloud Agents; dots can coordinate supported Work or Codex tasks.2, 8

Cursor Cloud Agents and Kiro cloud sessions using autonomous mode are an apples-to-apples comparison for delegated software work. Give them the same starting repository, issue and checks, then compare the patch and review effort. Those development environments serve a different task from keeping the weekly business brief moving.28, 29, 30

The product names can obscure that distinction. Grok Build is the dedicated coding agent with terminal, headless and Agent Client Protocol workflows. Claude Code and Codex also belong in the development discussion, while Cowork and dots serve the broader workflows described here.11, 31, 32 Our coding-agent comparison follows that decision in more detail.

What does it cost to keep the responsibility moving?

The useful budget is the one for the job you have now defined. Access to the coordinating agent is one part; connected services, delegated work and human review may add more. A lower subscription price helps only if the required workflow is available and produces a useful result.

These are selected routes checked on 11 November 2026. They do not have equivalent allowances, and account, regional or workspace terms can differ.

Scroll the table sideways on smaller screens.

Budget for the route and operating model you would actually use
ProductSelected access routeAdditional cost question
Grok BotThe X Subscription Premium is listed at US$30/month with Grok Bot included; other eligible and legacy routes are documented.6, 13Linked account, included usage and additional charges
OpenAI dotsEligible Pro and workspace routes, subject to rollout and region. OpenAI lists Pro options at US$100, US$200 and US$500/month.7, 12Actual dots eligibility and delegated Work or Codex usage
Claude CoworkPaid plans; Pro is listed at US$20 billed monthly.14, 22Plan allowance and desktop-dependent tasks
Gemini tasksEligible personal accounts with Google AI Pro or Ultra; check regional pricing.16, 23Account eligibility, usage and required connected actions
Amazon QuickFree route; Plus at US$20/user/month annually or US$25 monthly. Organisation plans add a US$250/account/month infrastructure fee to seat pricing.26Plan route, usage, overages and infrastructure charges
HermesOpen-source software with configured model access.20Inference, execution, hosting and maintenance
OpenClawOpen-source gateway and your configured runtime.18Providers, infrastructure, connections and operation
Kiro CrewOpen-source agent on the hardware and runtime you operate.27Runtime usage, infrastructure and maintenance

OpenAI specifies that tasks a dot starts or manages in Work or Codex count towards those products' usage limits.7 More generally, record the complete cost in RM after conversion, tax and added usage where applicable. Include time spent correcting the brief and keeping the connections working.

Where I would start

I would begin with the product that can reach the approved project information and return the first brief in a place the team will review. Between Grok Bot and dots, the next question is which organising model feels clearer: a Bot with a distinct role, or a personal coordinator carrying the responsibility across tasks.

Give it one real deadline, the approved sources and an explicit review point. After the first draft, change a requirement and return for the next run. Keep the resulting documents so you can compare what improved and what still needed correction. That trial tells you more about the fit than a broad promise of autonomy.

Cowork, Gemini tasks or Amazon Quick are worth including when their connections and deliverables fit the job better. Hermes, OpenClaw or Crew become relevant when control of the platform is worth operating it yourself. Expand the assignment once the first responsibility is working well enough to maintain.

Questions about access and capabilities

What is an AI coworker, and how is it different from a remote coding agent?

In this guide, an AI coworker handles ongoing work across information, apps and conversations. A remote coding agent works in a development environment and is evaluated on code changes, tests and reviewable results. Remote execution describes where work runs; it does not by itself make two products equivalent.

Sources behind this guidance: Grok Bot: overview · OpenAI: meet dots and check eligibility · Cursor: Cloud Agents · Kiro: cloud development sessions

Are Cursor Cloud Agents and AWS Kiro an apples-to-apples comparison?

Cursor Cloud Agents and Kiro cloud sessions with autonomous mode are a fair comparison for delegated software work. Give them the same repository state, task, checks and review criteria. Amazon Quick and Kiro Crew serve different roles, so identify the AWS product before comparing it with a general AI coworker.

Sources behind this guidance: Cursor: Cloud Agents · Kiro: cloud development sessions · Kiro: autonomous mode

Which AWS product is closest to Grok Bot and OpenAI dots?

Amazon Quick is the closer hosted business-work comparison. Kiro Crew is relevant when comparing persistent agents you operate yourself, such as Hermes or OpenClaw. Kiro cloud and autonomous mode belong in a remote-coding comparison. These are editorial categories based on the documented workflows.

Sources behind this guidance: AWS: Amazon Quick capabilities · Kiro Crew: persistent open-source agent · Kiro: cloud development sessions

What is the closest Anthropic alternative to Grok Bot and dots?

Claude Cowork is the relevant broader-work comparison. It supports files, connectors, skills and scheduled work. Claude Code is the development-focused comparison. Check current Cowork behaviour by plan and surface rather than relying on older descriptions of desktop-only execution.

Sources behind this guidance: Anthropic: Cowork on web, desktop and mobile · Anthropic: scheduled Cowork tasks · Anthropic: Claude Code overview

Is Gemini Spark still the name of Google’s agent?

Google’s current help page says Spark was renamed to Task. This guide calls the experience tasks in Gemini. It supports ongoing workflows and schedules, with account and subscription requirements. Google Workspace Studio is a separate product for building and sharing Workspace automations.

Official sources: Google: tasks in Gemini, eligibility and the Spark name change · Google: schedules for tasks in Gemini · Google: Workspace Studio for agent-based automations

Are Hermes and OpenClaw free alternatives to hosted AI coworkers?

They are open-source options you can configure and operate. The software licence cost is only one part of the budget: model access, hosting, connected services and maintenance can still cost money. Compare a working deployment with the hosted product, including who handles updates and failures.

Sources behind this guidance: Nous Research: Hermes Agent · OpenClaw: official documentation

Is Grok Bot the same as Grok Build?

No. Grok Bot is a broader AI teammate for work across apps and a persistent cloud computer. Grok Build is a coding agent with an interactive terminal interface, headless execution and Agent Client Protocol support. Choose the comparison around the work you need done.

Official sources: Grok Bot: overview · Grok Build: coding agent overview

Are OpenAI dots the same as Codex?

No. A dot coordinates ongoing responsibilities and can delegate or continue supported development tasks. Codex is the development workflow for building, debugging, testing and reviewing software. A dot can coordinate a coding task without being the same product as Codex.

Official sources: OpenAI dots: tasks and memory · OpenAI: Codex Cloud

Can Grok Bot and dots work while my laptop is closed?

Their cloud work can continue while your laptop is off. Work that uses your own connected computer still depends on that computer being available. For dots, the documentation specifies that the connected computer must be online with the ChatGPT app open for local work.

Official sources: Grok Bot: computer and apps · OpenAI dots: computers and apps

Will connecting an app automatically make the agent monitor it?

Access and an assigned workflow are separate. Specify what to check, which sources to use, the schedule or supported event, and when to notify you. Verify that the task or routine was saved. Do not treat a connected account as evidence that monitoring is active.

Sources behind this guidance: Grok Bot: skills and routines · OpenAI dots: tasks and memory

Which is better for a small business?

Start with the product that supports your exact apps, is available to your account and can produce a reviewable result for one recurring job. Compare output quality, corrections, review effort and actual usage cost. This documentation comparison does not establish a universal winner or measured business return.

Techsona Labs editorial guidance. Read the comparison method.

Sources and how this comparison was made

Documentation reviewed . This AI-assisted comparison uses the official product, help and pricing pages below. The weekly-brief example and recommendations are editorial guidance. No shared hands-on benchmark, measured ROI or universal performance ranking is claimed.

Official references

Each reference identifies the product facts it supports. Check the provider's current terms and your account before relying on a particular access route or action.

  1. Grok Bot: overview docs.x.ai The broader AI teammate role, conversations and work across apps and tools.
  2. Grok Bot: how it works, memory and coding handoff docs.x.ai Separate Bot roles and memories, shared account context and delegation to Cursor Cloud Agents.
  3. Grok Bot: skills and routines docs.x.ai Reusable procedures, scheduled or event-triggered work and validating a task before automating it.
  4. Grok Bot: computer and apps docs.x.ai The shared persistent cloud computer, files, sessions, credentials and app connections.
  5. Grok Bot: approvals, security and privacy docs.x.ai Auto Review and custom approval rules; the conditions around permitted actions.
  6. Grok Bot: getting started and access routes docs.x.ai Account linking, eligible subscriptions and the documented legacy access routes.
  7. OpenAI: meet dots and check eligibility learn.chatgpt.com The personal coordinating-agent role, eligibility and usage treatment for delegated Work or Codex tasks.
  8. OpenAI dots: tasks and memory learn.chatgpt.com Ongoing responsibilities, background tasks, saved notes, schedules and Work or Codex handoffs.
  9. OpenAI dots: computers and apps learn.chatgpt.com Cloud and connected computers, plugin access and availability requirements for local work.
  10. OpenAI dots: controls and stopping work learn.chatgpt.com Action review, custom rules and the differences between pausing, stopping tasks and cancelling schedules.
  11. Grok Build: coding agent overview docs.x.ai The separate coding product, its terminal interface, headless execution and Agent Client Protocol support.
  12. OpenAI: subscription pricing and usage learn.chatgpt.com The selected US-dollar Pro subscription prices; product eligibility is checked separately.
  13. Cursor / The X Subscription: pricing cursor.com The X Subscription Premium price and its listed Grok Bot inclusion.
  14. Anthropic: Cowork on web, desktop and mobile support.claude.com Cowork surfaces, cloud sessions by plan and the desktop dependency for local-folder access.
  15. Anthropic: scheduled Cowork tasks support.claude.com Recurring assignments and remote execution of scheduled Cowork work.
  16. Google: tasks in Gemini, eligibility and the Spark name change support.google.com The Spark-to-Task name change, supported workflows and personal-account eligibility requirements.
  17. Google: schedules for tasks in Gemini support.google.com Time-based, Gmail and topic monitors, including the limitation for time-critical alerts.
  18. OpenClaw: official documentation docs.openclaw.ai The self-hosted gateway model, channels, sessions, agent routing and runtime responsibilities.
  19. OpenClaw: automations and scheduling docs.openclaw.ai Persistent scheduled jobs and delivery to configured channels or webhooks.
  20. Nous Research: Hermes Agent hermes-agent.nousresearch.com The open-source assistant, persistent memory, skills and configurable execution environment.
  21. Hermes Agent: scheduled tasks hermes-agent.nousresearch.com One-off and recurring work, result delivery and the running-gateway requirement for agent jobs.
  22. Anthropic: plan pricing claude.com The selected monthly Pro price and other subscription routes.
  23. Google: AI subscription plans one.google.com Regional Google AI subscription offers; no universal numeric price is assumed here.
  24. Google: Workspace Studio for agent-based automations workspace.google.com Workspace Studio as a separate environment for building and sharing Workspace automations.
  25. AWS: Amazon Quick capabilities aws.amazon.com Business information, connected apps, research, deliverables and workflows. Prices use reference 26.
  26. AWS: current Quick pricing and account fees aws.amazon.com Free and Plus routes, billing cadence and infrastructure fees for organisation plans.
  27. Kiro Crew: persistent open-source agent kiro.dev Self-operated local or remote execution, memory, recurring jobs and messaging channels.
  28. Cursor: Cloud Agents cursor.com Remote development environments, delegated repository work and reviewable coding results.
  29. Kiro: cloud development sessions kiro.dev Remote coding sessions and their development environment.
  30. Kiro: autonomous mode kiro.dev Delegated development work used in our Cursor-versus-Kiro comparison.
  31. Anthropic: Claude Code overview code.claude.com Repository editing, commands and software-development workflows, distinguishing Code from Cowork.
  32. OpenAI: Codex Cloud learn.chatgpt.com Coding tasks, reusable development environments and inspecting code changes and check results.
Back to blog

Techsona Labs

Put an idea into practice

Have a workflow that needs a better system?