# AI agents and coworkers from the major AI makers, compared > Ten ready-made AI agents from the makers of the major AI assistants, compared by who can use them today, what the plans cost, whether they keep working when you close the app, what apps they reach, and when they ask before acting. - Page: https://allthingscompared.com/ai-agents-and-coworkers/ - JSON: https://allthingscompared.com/ai-agents-and-coworkers.json - Markdown: https://allthingscompared.com/ai-agents-and-coworkers.md - 10 options compared. Facts checked Nov 6, 2026. Page updated Nov 6, 2026. - Order: alphabetical by name; nothing is featured. ## How to read this - **Verified**: seen on the provider's own page on the date shown. The link goes to that page. - **Provider claim**: something the provider says about itself, not confirmed by us. - **Not publicly stated**: the provider's pages do not say. That is not the same as "no". - **"Our read" lines**: our interpretation of what a provider publishes. They are not scores, ratings or endorsements. ## Needs - **Open to one person for $50 a month or less**: The maker's pages list a plan or free tier that includes the agent, that one person can buy, at $50 a month or less (an annual-billing price counts by its per-month figure, and the billing term is stated on the option). - **Keeps working after you close the app or laptop**: The maker's pages say the agent keeps working after you close the app, close your laptop or turn off your device, and do not say that its scheduled work stops when the device is off. These are the maker's own statements. - **Has its own computer or email address**: The maker's pages say the agent has a computer or browser of its own, or its own email address, account or calendar (a shared cloud sandbox or a chat handle does not count). - **Asks before sensitive actions**: The maker's pages say the agent asks for approval or confirmation before certain actions, such as sending messages, deleting files or paying. This is the maker's own statement about its design. - **A free plan includes it**: The maker's pages list a $0 plan with no end date that includes the agent, with limited use (a trial or a one-time allowance does not count). ## Options ### ChatGPT Work OpenAI's agent inside ChatGPT for long tasks and finished files such as sheets, slides, docs and web apps, with scheduled and event-triggered tasks. A limited desktop version is on the Free plan. - Website: https://openai.com/index/chatgpt-for-your-most-ambitious-work/ - Based: Online - Card on the page: https://allthingscompared.com/ai-agents-and-coworkers/#chatgpt-work - Our read (interpretation, not a fact): People who already use ChatGPT and want an agent that works across their apps and files, from a limited free version up to Pro. - Fits: Open to one person for $50 a month or less; Keeps working after you close the app or laptop; Asks before sensitive actions; A free plan includes it - **Who can use it today:** OpenAI's pricing page lists ChatGPT Work as 'Limited (desktop app)' on Free and Go, and as 'Desktop, web, and mobile' on Plus and Pro. A release note (July 9, 2026) says 'Work is available in all supported regions.' At launch OpenAI said web and mobile were 'rolling out today' for Pro, Enterprise and Edu and would reach Plus and Business 'over the next few days'; the pricing page now lists Plus with web and mobile. Enterprise and Edu workspaces had Work off by default during a preview. Pricing page is undated; it was read in the browser pane because plain fetches of chatgpt.com return 403. The launch wording is from the launch post (July 9, 2026) and the release notes (https://help.openai.com/en/articles/6825453-chatgpt-release-notes), read 2026-11-05. The Business plan page lists 'All ChatGPT, ChatGPT Work, and Codex features'. [Verified · Nov 6, 2026](https://chatgpt.com/pricing/) - **Lowest-priced plan that includes it:** $0 per month (Free plan: 'Limited ChatGPT Work access on desktop'). Plus is $20 per month (US dollars, monthly billing) with 'Expanded access to ChatGPT Work on desktop, web, and mobile'. Go is $8 per month with limited desktop access. Pro is from $100 per month ($100, $200 or $500 by usage tier) with 'Longer Codex and ChatGPT Work sessions'. The page says Go 'may include ads'. Business Standard seats are $20 per seat per month billed annually or $25 billed monthly (https://chatgpt.com/pricing/?type=team, read 2026-11-05). The three Pro tier prices are from https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers, read 2026-11-05. [Verified · Nov 6, 2026](https://chatgpt.com/pricing/) - **What it does:** OpenAI's launch post says 'ChatGPT Work is an agent that can take action across your apps and files, stay with a project for hours if needed, and turn a goal into finished work.' It names finished sheets, slides, docs and web apps; examples include month-end variance analysis, campaign briefs and sales-meeting prep, and (release note, August 25, 2026) filling in forms on signed-in websites such as a DMV appointment or a passport renewal. OpenAI's older ChatGPT agent is retired: its help article says 'ChatGPT agent is no longer available. Use ChatGPT Work for longer, multi-step tasks and finished deliverables.' [Verified · Nov 5, 2026](https://openai.com/index/chatgpt-for-your-most-ambitious-work/) - **Background and scheduled work:** ChatGPT Work states that it can 'stay with a project for hours' and that 'Even when you're away from your computer or phone, ChatGPT Work can keep projects moving forward with Scheduled Tasks.' OpenAI's docs say cloud runs suit scheduled tasks that research or check websites 'because your computer doesn't need to be awake'. A release note (June 17, 2026) says tasks 'cannot run more than once per hour, and unattended tasks may automatically pause after a period of inactivity'. Event triggers (Gmail, Slack, GitHub) are on Plus and above, not on Free or Go. The cloud statement is in the docs at https://learn.chatgpt.com/docs/get-started-with-work; the event-trigger plans are from the August 25, 2026 release note. [Provider claim · Nov 5, 2026](https://openai.com/index/chatgpt-for-your-most-ambitious-work/) - **Apps and connectors:** OpenAI's launch post says plugins connect ChatGPT Work to 'tools your team uses, like Slack, Google Drive, SharePoint, email, calendars, customer relationship management systems, and project trackers'. Its docs list featured plugins including Google Drive, Outlook Email, Gmail, Teams, Slack, SharePoint, Salesforce, Adobe Acrobat, Zoom, LinkedIn, Google Calendar, GitHub, Canva and Dropbox, and there is a Plugin Directory. On desktop it also has a built-in browser and Computer Use. The Work pages read give no count of plugins. OpenAI's launch post for Dots says its plugin ecosystem reaches 'over 4,000 apps'; that figure is on the Dots option, not on a Work page. [Verified · Nov 5, 2026](https://learn.chatgpt.com/docs/get-started-with-work) - **Own identity and access:** ChatGPT Work acts with your own accounts and permissions: on desktop, 'Computer Use lets ChatGPT use your computer on your behalf', and plugins use your connected accounts. The pages read describe no own email address, account or calendar for ChatGPT Work. OpenAI's separate Dots product has its own cloud computer; see that option. [Verified · Nov 5, 2026](https://openai.com/index/chatgpt-for-your-most-ambitious-work/) - **Approvals and controls:** ChatGPT Work states that you 'can follow its progress, answer questions, change direction, and approve important actions as it works', and a release note (August 25, 2026) says it 'will always ask for confirmation before consequential actions, such as completing a reservation or payment'. OpenAI's launch post adds 'You decide what it can access, when it should check in, and when it needs your approval before taking action.' Statements are OpenAI's own and were not tested. Admin controls and audit logs for Enterprise are listed in the pricing docs. [Provider claim · Nov 5, 2026](https://openai.com/index/chatgpt-for-your-most-ambitious-work/) - **Models named:** OpenAI's docs say to 'Choose GPT-6 Astra for demanding work' and name GPT-6.1 Sol, GPT-6.1 Terra and GPT-6 Luna for lighter tasks. The launch post (July 9, 2026) named GPT-5.6. A release note (October 26, 2026) adds GPT-6.1 Terra to Work and Codex. Model names on OpenAI's pages change often. [Verified · Nov 5, 2026](https://learn.chatgpt.com/docs/get-started-with-work) - **Usage limits:** OpenAI's pricing docs say 'ChatGPT Work and Codex share usage.' For Plus and Business Standard they give estimates of local messages per five-hour period, for example 5 to 45 on GPT-6 Astra and 350 to 3,000 on GPT-6 Luna; 'Pro plans currently have no five-hour limit', weekly limits may also apply, and credits can be bought. Free and Go list limited access with no numbers. Docs page is undated. The estimates are OpenAI's and depend on task size. [Verified · Nov 5, 2026](https://learn.chatgpt.com/docs/pricing) - **Data and privacy:** ChatGPT Work states on OpenAI's pricing page 'Content is used to train our models: Opt-out available' for Free, Go, Plus and Pro, and 'No training on your business data by default' for Business. A release note (August 25, 2026) states that for signed-in websites 'ChatGPT cannot see your username or password, and they are never seen by the model or used in model training'. The Business wording is from https://chatgpt.com/pricing/?type=team and the sign-in statement from the release notes, both read 2026-11-05. OpenAI's data controls article describes where the training setting is. [Provider claim · Nov 6, 2026](https://chatgpt.com/pricing/) - **Support:** OpenAI's help centre offers AI chat and a phone line (1-888-GPT-0090; 'Calls may be recorded to improve OpenAI services'). Support hours are not stated. The footer line is shared by all ChatGPT help articles; no ChatGPT Work specific support terms were found. [Verified · Nov 5, 2026](https://help.openai.com/en/articles/20001530-getting-started-with-your-dot) - Not clear from public sources: OpenAI's pages give no usage numbers for Free or Go and show Plus and Business limits only as estimates. They describe no own email address or calendar for the agent. The pricing page is undated, and its scheduled-tasks row and OpenAI's help article differ on whether Free can have scheduled tasks. ### Claude Cowork Anthropic's agent for knowledge work in the Claude apps: it works in folders and tools you choose, runs sessions and scheduled tasks in the cloud, and is included in paid Claude plans from Pro. - Website: https://claude.com/product/cowork - Based: Online - Card on the page: https://allthingscompared.com/ai-agents-and-coworkers/#claude-cowork - Our read (interpretation, not a fact): People on a paid Claude plan who want a task agent that works on their files and connected tools. - Fits: Open to one person for $50 a month or less; Keeps working after you close the app or laptop; Asks before sensitive actions - **Who can use it today:** Anthropic's help article says 'Claude Cowork is available on paid plans (Pro, Max, Ultra, Team, and Enterprise)'. The desktop app for macOS, Windows and Linux (beta) is 'Available on all paid plans'; web and mobile are listed for Pro, Max, Ultra and Team (in beta on Team) and, for Enterprise, 'when an owner enables it'. It is not on the Free plan. The product page says 'Works with a paid plan. Rolling out on web, mobile.' and the help article says the combined chat-and-Cowork experience 'is rolling out to Pro, Max, and Ultra plans'. No region limit is stated on the pages read. The help article shows 'Updated this week'. The product page (https://claude.com/product/cowork) is undated; its newest dated item is August 26, 2026. [Verified · Nov 6, 2026](https://support.claude.com/en/articles/13345190-get-started-with-claude-cowork) - **Lowest-priced plan that includes it:** $20 per month billed monthly, or $17 per month with an annual subscription ($200 billed up front) (Pro plan, per person, US dollars, before tax). The product page says 'Claude Cowork is included in your Pro plan' and that it 'consumes limits faster than Chat'. Max is $100 per month and Ultra $200 per month. Team seats start at $20 per seat per month for teams of 2 to 150. Anthropic's pricing page (https://claude.com/pricing, read 2026-11-05) lists Max and Ultra as billed monthly and Team seats at $20 billed annually or $25 billed monthly. That page's individual feature list does not mention Cowork, while its Team seat list says 'Claude Code and Claude Cowork'. Prices exclude tax and are 'subject to change at Anthropic's discretion'. [Verified · Nov 6, 2026](https://claude.com/product/cowork) - **What it does:** Anthropic's product page says 'Claude Cowork completes tasks you can steer from anywhere. Give it a goal, and it works across your files and tools.' It 'works directly in folders and tools you choose, and runs your task to deliver work for review.' The help article says 'Claude can take on complex, multi-step tasks and execute them on your behalf.' Examples on the pages: organize a Downloads folder, turn receipts into an expense report, build spreadsheets with working formulas and slide decks, and a Monday readout that pulls web views and signups from Amplitude. A built-in browser opens in the Cowork side panel. Computer use is labelled 'in research preview for both Cowork and Claude Code' on the product page FAQ. [Verified · Nov 6, 2026](https://claude.com/product/cowork) - **Background and scheduled work:** Claude Cowork states 'Close your laptop, it keeps going. Schedule a task for any cadence, and it runs unattended.' Anthropic's help article says 'In sessions in the cloud, Claude keeps working when you close your laptop or step away' and 'Scheduled tasks run in the cloud, with no device online'. Tasks that use local files, your browser or your computer need the desktop app left open. The pages state no limit on how long a task runs ('Tasks can run for extended periods depending on complexity'). The first sentence is on the product page. The scheduled-task setup is described in https://support.claude.com/en/articles/13854387-schedule-recurring-tasks-in-claude-cowork. [Provider claim · Nov 6, 2026](https://support.claude.com/en/articles/13345190-get-started-with-claude-cowork) - **Apps and connectors:** Anthropic's product page shows connectors for Amplitude and Microsoft 365 and says Claude Cowork and the Slack connector are included in Team plans. It describes plugins that 'bundle your tools, knowledge, and workflows into a single install' and sub-agents. Anthropic's pages describe a connectors-and-plugins directory and say Claude 'reaches for your connectors and integrations first, falls back to your browser when needed, and only uses your screen as a last resort.' The pages give different counts of connectors, so no number is given here. Anthropic's marketplace page (read through a summarizing tool) showed 1,079 connectors, and a Claude Duet page says 'more than 6,000 connectors and plugins'; the counts are live and differ. [Verified · Nov 6, 2026](https://claude.com/product/cowork) - **Own identity and access:** Anthropic's help article says 'Claude has access to the local files you grant it permission to access, and can take real actions on your behalf.' Claude can only read and write files in folders you connect, and cloud sessions run 'in an isolated environment on Anthropic's servers'. The pages read describe no own email address, account or calendar for Cowork. On Team and Enterprise plans an owner turns Cowork on or off for the organization. [Verified · Nov 5, 2026](https://support.claude.com/en/articles/13345190-get-started-with-claude-cowork) - **Approvals and controls:** Claude Cowork states that it has three approval modes: Manual (default on Enterprise plans), Auto (default on Pro, Max, Ultra and Team plans), in which Claude 'reviews each action for safety' and blocks what it judges unsafe, and Skip, where nothing is checked automatically. Anthropic states 'When using Cowork, Claude requires your explicit permission before permanently deleting any files.' The product page says 'Claude asks before acting' and, with permissions settings enabled, that Claude 'waits for your approval before anything significant'. Team and Enterprise admins can set access by team, manage spend limits and stream activity to a SIEM. Anthropic also says no approval mode replaces your judgment. The admin sentence is from the product page. The default on individual plans is Auto, where Claude's safety review rather than you decides some actions. [Provider claim · Nov 6, 2026](https://support.claude.com/en/articles/13345190-get-started-with-claude-cowork) - **Models named:** Not publicly stated. Anthropic's Cowork product page, help article and web, desktop and mobile article name no model for Cowork. The pricing page names Opus, Sonnet, Haiku and Fable for Claude plans in general. - **Usage limits:** Anthropic's pricing page says 'Usage limits apply' and the product page says Cowork 'consumes limits faster than Chat'. Activity across Claude on web, desktop, mobile and Claude Code 'all draws from the same pool'. Max gives '5x Pro usage every 5 hours and every week' and Ultra '10x Pro usage every week' with no 5-hour limit for verified accounts. Anthropic says it may limit usage 'in other ways, such as weekly and monthly caps'; paid plans can turn on usage credits to keep working at API rates. No message counts are given for any plan. [Verified · Nov 5, 2026](https://claude.com/pricing) - **Data and privacy:** Claude Cowork states in the model-training row of Anthropic's pricing page 'Opt-out' for Free, Pro, Max and Ultra and 'None by default' for Team and Enterprise. Anthropic's help article says a deleted task is removed 'from our backend storage systems within 30 days'. The Cowork pages read state no other retention period. Anthropic's Claude Code data page gives separate retention figures for that product; the privacy centre was not read for Cowork. [Provider claim · Nov 5, 2026](https://claude.com/pricing) - **Support:** Anthropic's help article says 'Support is provided in writing through the support messenger and email. We don't offer phone support.' Hours are not stated ('Response times vary by plan and by the severity of the issue'); organization owners and admins on qualifying Enterprise plans 'have prioritized options in the messenger'. The article shows 'Updated this week'. [Verified · Nov 6, 2026](https://support.claude.com/en/articles/9015913-how-to-get-support) - Not clear from public sources: Anthropic's Cowork pages name no model for Cowork and state no region limit. They use 'rolling out' for web and mobile and for the merged chat-and-Cowork experience. The pricing page's individual-plan feature list does not mention Cowork, while the product page and help article say it is included from Pro. On Pro the default approval mode is Auto, where a safety check decides some actions. ### Dots in ChatGPT OpenAI's always-on agent in ChatGPT: each user can create one dot that has its own cloud computer and browser, keeps working between conversations, and can be reached in ChatGPT, Slack or Teams. It is included in Pro and Business Premium plans. - Website: https://chatgpt.com/features/dots/ - Based: Online - Card on the page: https://allthingscompared.com/ai-agents-and-coworkers/#dots-in-chatgpt - Our read (interpretation, not a fact): People who can pay for a Pro plan or a Business Premium seat and want an agent that keeps making progress on ongoing responsibilities. - Fits: Keeps working after you close the app or laptop; Has its own computer or email address; Asks before sensitive actions - **Who can use it today:** OpenAI's dots page says 'Dots are rolling out in ChatGPT on web, mobile, and desktop. Access is available now across Pro, Business Premium, and Enterprise plans in eligible markets.' The help article says dots are 'rolling out in ChatGPT to Pro users aged 18 and older in markets excluding the European Economic Area, Switzerland, and the UK', are 'also available to Business Premium users in supported ChatGPT regions', and that 'For personal and Business Premium accounts, dots are rolling out gradually. Access may take several days to reach your account.' Not on Free, Go or Plus. Enterprise, Edu and Healthcare: 'generally available', with an admin preview from November 4 through November 18, 2026. Texting a dot is 'available in beta to a limited number of Pro users in the US'. Each member can create one personal dot. Help article (https://help.openai.com/en/articles/20001530-getting-started-with-your-dot) shows 'Updated: yesterday'. No waitlist or request-access wording was found on the Pro path; a 'Contact sales' button on the dots page relates to specialist dots. The launch post is dated September 29, 2026. Both pages were read in the browser pane because plain fetches of chatgpt.com and help.openai.com return 403. [Verified · Nov 6, 2026](https://chatgpt.com/features/dots/) - **Lowest-priced plan that includes it:** Pro, from $100 per month (US dollars, billed monthly; three usage tiers at $100, $200 and $500 per month). OpenAI states 'Your first dot is included in your Pro or Business Premium plan at no extra cost.' A Business Premium seat is $100 per seat per month billed annually ($125 billed monthly), for teams of 2 or more. Go ($8 per month) and Plus ($20 per month) list Dots as 'No'. Pricing page is undated; its Pro card reads 'Dot, your always-on agent', and its compare table marks Dots 'No' for Free, Go and Plus (marks read 2026-11-05). The 'no extra cost' sentence is from the launch post (https://openai.com/index/introducing-dots/), the billing-monthly-only statement from https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers, and the Business Premium seat price from https://chatgpt.com/pricing/?type=team, all read 2026-11-05. The help article says dots are 'free to preview' for Enterprise workspaces from November 4 through November 18, 2026, then billed. [Verified · Nov 6, 2026](https://chatgpt.com/pricing/) - **What it does:** OpenAI's dots page says a dot 'uses its own computer to move your projects forward' and that you 'hand over a big project or something smaller'. Examples on the page: a dot that tracks product usage and revenue and refreshes an investor deck, checks customer requirements against emails and product docs, watches customer feedback and builds and tests fixes for review, or drafts show notes and social posts for approval. The help article says it can 'research, create, and follow through on tasks'. OpenAI's launch post (September 29, 2026, read 2026-11-05) says dots 'can do nearly anything using their own cloud computer, their own browser, and the apps you've connected'; that is OpenAI's own wording. [Verified · Nov 6, 2026](https://chatgpt.com/features/dots/) - **Background and scheduled work:** Dots in ChatGPT states 'Your dot lives in the cloud and has its own computer and browser. You can reach it and it can keep working even when your computer is off.' The help article says a dot 'can research in the background', run reminders and recurring checks you schedule, and keep making progress between conversations. Pausing a dot 'does not delete its information', and separately scheduled tasks may continue. The pages read state no limit on how long a task runs. Local-computer access is optional and starts off; the help article says to 'Keep that computer online with the ChatGPT app open' for work on it. [Provider claim · Nov 6, 2026](https://learn.chatgpt.com/docs/dots) - **Apps and connectors:** Dots in ChatGPT states that 'Through our ecosystem of plugins, they can readily connect to over 4,000 apps' (launch post, September 29, 2026). The help article says apps are connected through Plugins, a dot can be added to Slack and Microsoft Teams, can use Codex cloud environments and, if you allow it, your own computer. Plugins named in OpenAI's examples include Gmail, Google Drive and GitHub. The count is OpenAI's own figure for its plugin ecosystem. Help article read 2026-11-06 says 'Available apps may differ'. [Provider claim · Nov 5, 2026](https://openai.com/index/introducing-dots/) - **Own identity and access:** OpenAI's pages say a dot has 'its own cloud computer' and browser and a handle ('Its default handle is @yourname-dot'). On email the help article says 'At launch, you cannot give your dot its own standalone email address'; you can connect your personal email account for it to use. 'Your dot cannot initiate calls to you at launch.' Its cloud browser 'has its own sign-in sessions'. OpenAI describes 'specialist dots' with their own identity only for enterprise pilots (set aside). The existing ChatGPT app permissions apply to a dot. [Verified · Nov 6, 2026](https://help.openai.com/en/articles/20001530-getting-started-with-your-dot) - **Approvals and controls:** Dots in ChatGPT states 'Set Custom Rules to allow specific actions, require approval, or block them.' It also states 'Actions that could affect your accounts or share information go through Auto-review for checks against your instructions, Custom Rules, and safety requirements. Some need your approval. Steps like changing your password stay with you.' The help article lists rule behaviors: take action without asking, take action if pre-approved, ask before taking action, and hand off to you. You can pause a dot or take over its computer; OpenAI notes 'Your dot can make mistakes'. Statements are OpenAI's own and were not tested. Workspace admins can pause dots and restrict access. [Provider claim · Nov 6, 2026](https://chatgpt.com/features/dots/) - **Models named:** OpenAI's dots page says 'Dots run on OpenAI's family of frontier models'. The docs say the models that power dots now include GPT-6.1 Terra and that 'Your dot can draw on different models for different parts of its work'. The launch post said 'Powered by GPT-6 Astra'. The launch post carries an update dated November 3, 2026 adding GPT-6.1 Terra. [Verified · Nov 6, 2026](https://learn.chatgpt.com/docs/dots) - **Usage limits:** OpenAI's pages state no number for dots. Conversations with your dot 'don't count toward your ChatGPT usage limits', while work it hands to ChatGPT Work or Codex counts as usual. The launch post says the plan 'includes an allowance for deeper work, with extended limits for the first month after launch'. Each member can create one personal dot. The pages do not say whether the first-month extended limits have ended. Pro plans list no five-hour limit in ChatGPT Work and Codex. [Verified · Nov 5, 2026](https://openai.com/index/introducing-dots/) - **Data and privacy:** Dots in ChatGPT states 'On personal ChatGPT plans, you can control whether dots' conversations and work are used to improve our models', that OpenAI does not use content from Business, Enterprise or Edu workspaces to improve its models by default, and that it does not 'train directly on proactive research or your dot's notes to itself' (launch post, September 29, 2026). The pricing page lists 'Opt-out available' for training on Pro. The help article says texts to a dot go through a third-party provider and 'the conversation is not end-to-end encrypted between you and your dot'. The pricing-page row and the help-article statement were read 2026-11-06. [Provider claim · Nov 5, 2026](https://openai.com/index/introducing-dots/) - **Support:** OpenAI's help centre offers AI chat and a phone line (1-888-GPT-0090; calls may be recorded). Support hours are not stated. For bugs or feedback OpenAI says to ask your dot in its main conversation. The footer is shared across ChatGPT help articles. [Verified · Nov 5, 2026](https://help.openai.com/en/articles/20001530-getting-started-with-your-dot) - Not clear from public sources: OpenAI's pages mix 'available now' with 'rolling out gradually' and say access 'may take several days to reach your account'. They give no number for a dot's usage allowance, list no standalone email address for a dot at launch, and describe enterprise 'specialist dots' that are pilots only. The launch post calls the Enterprise offer a beta; the help article calls it generally available. ### Grok Bot SpaceXSI's agent that works on a persistent cloud computer of its own, with routines on a schedule or on events, connected apps and an optional email address. It is included in The X Subscription from the Premium plan. - Website: https://x.ai/bot - Based: Online - Card on the page: https://allthingscompared.com/ai-agents-and-coworkers/#grok-bot - Our read (interpretation, not a fact): People who want Bots that run routines on a schedule or on events and work in their apps from desktop or phone. - Fits: Open to one person for $50 a month or less; Keeps working after you close the app or laptop; Has its own computer or email address; Asks before sensitive actions - **Who can use it today:** SpaceXSI's Grok Bot page says 'Grok Bot is included with Premium, Super, Ultra, Teams, and Enterprise' and that it is 'generally available for Enterprise, as well as for teams on the Teams plan'. Its docs say it runs on macOS, Windows, Linux, iOS and Android. The Plus plan does not include it. No beta label appears on the current pages. Regions are not stated; the docs say tagging @bot on X 'is not available in every region yet'. The launch post (https://x.ai/news/introducing-grok-bot, August 11, 2026) said 'Grok Bot is in beta' and that Enterprise users 'can join a waitlist'; the current x.ai/bot FAQ says Enterprise access is by 'Contact sales'. The Plus exclusion is from https://docs.x.ai/x-subscription/grok-bot, read 2026-11-05. x.ai pages were read in the browser pane. [Verified · Nov 6, 2026](https://x.ai/bot) - **Lowest-priced plan that includes it:** Premium, $30 per month billed monthly (The X Subscription, US dollars, before tax). It includes 'Grok Bot's own computer', 'Routines on a schedule' and 'Weekly Grok Bot usage included'. Super is $100 per month (3.5x Premium usage) and Ultra $200 per month (12x). Teams start at $30 per user per month. Plus ($8 per month) does not include Grok Bot. SpaceXSI's docs show annual Premium billing at $300 per year (https://docs.x.ai/x-subscription/pricing, read 2026-11-05). Prices exclude applicable taxes. [Verified · Nov 6, 2026](https://x.ai/bot) - **What it does:** SpaceXSI's page says 'Bots have their own computer, so they can work inside your apps and tools.' Example Bots on the page: Inbox Manager, Account Manager, Talent Scout, Expense Manager, and a Sales Outbound Bot that 'researches accounts, scores contacts with intent, drafts email and LinkedIn in your voice, and leaves a review list for you to approve'. The launch post lists a sales Bot updating a CRM from call notes, an operations Bot processing invoices received in Gmail, and an engineering Bot reproducing a bug and filing a ticket. The launch-post examples are from https://x.ai/news/introducing-grok-bot, read 2026-11-05. [Verified · Nov 6, 2026](https://x.ai/bot) - **Background and scheduled work:** Grok Bot states that Bots 'run in parallel, 24/7, even when your laptop is closed'. The docs say 'Closing the app, laptop, or phone does not stop a background turn or routine.' Routines run on a schedule or on an event such as a Slack message, a GitHub or Linear event, an email or a webhook; a Bot 'can own up to 50 routines' and schedules must be at least five minutes apart. No overall task-duration limit is stated. The first sentence is on https://x.ai/bot. The routine counts are from https://docs.x.ai/grok-bot/skills-routines-and-automations, read 2026-11-05. [Provider claim · Nov 6, 2026](https://docs.x.ai/grok-bot/faq) - **Apps and connectors:** SpaceXSI's docs say Bots use 'a connected-app catalog' you browse and search, and can add an MCP server 'by URL or by a command that runs on its computer'. Apps named in the docs include Gmail, Notion, Slack, Google Calendar, Google Sheets, Google Drive, Google Docs, Google Slides, X, Zoom, SharePoint, Linear, 1Password and Outlook. Apps are account-wide, so any of your Bots can use a connected app. No total number of apps is stated on the pages read. x.ai also has a template marketplace for Bots (https://x.ai/bot/marketplace, not read in detail). [Verified · Nov 6, 2026](https://docs.x.ai/grok-bot/computer-and-apps) - **Own identity and access:** SpaceXSI's docs say Bots work from 'a persistent cloud computer with a browser, command line, and filesystem' that belongs to your account and is shared by all your Bots. The docs also say 'You can claim one email address at mail.grokbot.com' for your Bots to use; claiming it is optional, the address belongs to you rather than to one Bot, and on a team an admin must turn on Agent Email first. No own calendar or phone number is described. Bots sign in to other tools through the apps you authorize. Docs pages are undated. [Verified · Nov 6, 2026](https://docs.x.ai/grok-bot/computer-and-apps) - **Approvals and controls:** Grok Bot states 'Auto Review checks each action your Bot is about to take'; most actions run on their own, and when a check calls for it 'the Bot shows an approval card in your chat with the proposed operation and its inputs.' 'An action that matches one of your Ask first rules always gets a card.' A card raised by background work expires after 30 minutes, and email from a routine 'waits as a draft until you press Send.' Statements are SpaceXSI's own and were not tested. Enterprise access, network and audit controls are listed on the pricing comparison. [Provider claim · Nov 6, 2026](https://docs.x.ai/grok-bot/approvals-security-and-privacy) - **Models named:** SpaceXSI's docs say you cannot choose the model: 'Grok Bot picks the model for each task', which can be a SpaceXSI model or a third-party model from its sub-processor list. Grok 4.7 is the current SpaceXSI model across other products but the Grok Bot pages read do not tie it to Bot. [Verified · Nov 5, 2026](https://docs.x.ai/grok-bot/faq) - **Usage limits:** SpaceXSI's FAQ says 'Grok Bot subscriptions come with weekly usage included, with additional usage billed based on token cost.' The docs say 'Grok Bot draws from the same weekly usage pool as Cursor and the Grok apps' on The X Subscription, with prepaid credit covering usage beyond the pool. Super gives 3.5x and Ultra 12x Premium usage; no amount is stated for Premium. The launch post said Grok Bot 'comes with its own usage, separate from your Grok and Cursor plans'; the current docs describe the shared pool. [Verified · Nov 6, 2026](https://docs.x.ai/grok-bot/faq) - **Data and privacy:** Grok Bot states 'Grok Bot requires data storage and does not support Legacy Privacy Mode' and 'Training opt-out follows the applicable Cursor account and privacy settings.' The x.ai/bot FAQ says the cloud computer is encrypted in transit and at rest, with training opt-out. Statements are SpaceXSI's own. SpaceXSI's privacy policy (https://x.ai/legal/privacy-policy) was not read for this product. [Provider claim · Nov 6, 2026](https://docs.x.ai/grok-bot/approvals-security-and-privacy) - **Support:** SpaceXSI's docs footer lists email support (support@x.ai), a Discord community and a troubleshooting page. 'Priority support' is shown only for Enterprise on the pricing comparison. Support hours are not stated. Read through the docs pages and https://x.ai/pricing on 2026-11-05. [Verified · Nov 5, 2026](https://docs.x.ai/grok-bot/faq) - Not clear from public sources: SpaceXSI's pages state no number of connected apps and no weekly usage amount, say Grok Bot picks the model for each task, and give no support hours. The launch post (August 11, 2026) used the word 'beta' and an Enterprise waitlist; the current pages do not. The pages changed from separate Bot usage to a shared weekly pool. ### Meta Muse Meta's personal agent, free with a usage limit, with Power ($20 a month) and Maximum ($100 a month) plans. It works in the background, connects to apps, browses and shops, and asks before sending, posting or spending. - Website: https://ai.meta.com/muse/ - Based: Online - Card on the page: https://allthingscompared.com/ai-agents-and-coworkers/#meta-muse - Our read (interpretation, not a fact): People in the US or Canada who want a free-to-start personal agent on phone, Mac or the web. - Fits: Open to one person for $50 a month or less; Keeps working after you close the app or laptop; Asks before sensitive actions; A free plan includes it - **Who can use it today:** Meta's small-business page says 'Available today to people in the US and Canada who are 18 and over. We're working to bring Muse to more markets, and we're taking a measured approach in each one.' Its Muse FAQ says 'Muse is available for free with a usage limit.' Meta's launch post said Muse 'is rolling out in the US on iOS, Android, and muse.ai'. A Meta help article says 'Muse and Muse subscriptions are in limited testing and aren't available in all locations yet.' Muse runs on iPhone, Android, Mac and the web, and in WhatsApp. The FAQ is on https://ai.meta.com/muse/ and the limited-testing line on https://www.meta.com/help/subscriptions/1021145227643680/ (shows 'Updated: 8 weeks ago'), both read 2026-11-06. The launch post (published 2026-09-08, https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/) was read 2026-11-05. No waitlist or request-access wording was found. [Verified · Nov 6, 2026](https://muse.ai/business) - **Lowest-priced plan that includes it:** Free, with a usage limit (the weekly free allowance is not stated). Paid plans are the Power plan at $20 per month ('500M Muse tokens per week') and the Maximum plan at $100 per month ('3B Muse tokens per week'), US dollars, billed monthly and renewed automatically 'unless you cancel your subscription'. Extra tokens can be bought one time. The help article shows 'Updated: 8 weeks ago'. Meta says 'Subscription benefits and availability may vary by region and account.' Muse is not part of the Meta One plans described in Meta's September 15, 2026 post. [Verified · Nov 6, 2026](https://www.meta.com/help/subscriptions/1021145227643680/) - **What it does:** Meta's page says 'Muse is a personal AI agent that doesn't just answer questions, it does the work.' It says 'Tell Muse your goal once, and it keeps working in the background' and gives examples: update a grocery list, draft replies, build a client pitch, spin up a product page, plan a month of content. Meta's FAQ adds that Muse can browse the web, make purchases, create documents, set reminders and track goals. Meta's launch post says Muse 'can open a browser, fill out forms, and negotiate on their behalf'. The browsing and purchases wording is from the Muse FAQ on https://ai.meta.com/muse/. [Verified · Nov 6, 2026](https://muse.ai/business) - **Background and scheduled work:** Meta Muse states in its FAQ 'When you ask Muse to do something, it keeps working in the background whether you're actively using the app or not', and on its small-business page that 'it keeps working across the apps and services you connect, even after you close it.' Meta's help pages describe recurring tasks on a schedule, such as daily check-ins or weekly summaries, that continue until you cancel them. No task time limit is stated. The scheduled-tasks help article is https://www.meta.com/help/superintelligence/1484325780075655/ (read 2026-11-05, shows '8 weeks ago'). [Provider claim · Nov 6, 2026](https://ai.meta.com/muse/) - **Apps and connectors:** Meta's small-business page names Asana, Box, Canva, Dropbox, Figma, Granola, HighLevel, Intuit QuickBooks, Klaviyo, Lovable, Notion, Shopify, Slack, Stripe and Zoom, plus Facebook, Instagram and Meta ads, and says 'You can see the full list of available connectors in settings, with more to come.' Meta's help pages say you can ask Muse to create a custom connector, and that Meta does not review custom connectors. No total number of connectors is stated. Custom connector wording is from Meta's connectors help article (https://www.meta.com/help/superintelligence/1687253048996149/), read 2026-11-05. [Verified · Nov 6, 2026](https://muse.ai/business) - **Own identity and access:** Meta's Muse page describes 'A secure computer with a browser shared by you and your agent'; Meta's pages say each user shares a cloud computer with their Muse. Muse acts through the apps you connect and the permissions you set. Meta's Connect recap says Muse 'will have its own email address' (future wording); no own account, calendar or phone number is described as available today. The recap (https://www.meta.com/blog/meta-connect-2026-everything-we-announced/) was read 2026-11-05. Because the cloud computer is described as shared with you and the email address as future, this option is not tagged for having its own computer or email. [Verified · Nov 6, 2026](https://ai.meta.com/muse/) - **Approvals and controls:** Meta Muse states 'Review and approve actions from Muse before they happen, like sending emails and making purchases. See a complete audit trail of everything your agent has done (and what it is planning to do).' Its FAQ says Muse 'is designed to ask your permission before taking certain actions, like sending messages, making purchases, or sharing information with a connected app', and you can allow an action once, always allow it, or deny it. The small-business page says 'Nothing publishes, sends or spends without your approval, and you can undo it.' Statements are Meta's own and were not tested. Muse is a personal product; no admin controls are described. [Provider claim · Nov 6, 2026](https://ai.meta.com/muse/) - **Models named:** Meta's launch post says Muse is 'powered by Muse Spark, Meta's most capable model to date'. A later Meta post (October 12, 2026) says Muse Wonder 'will soon power more of Muse, rolling out first to subscribers'. The October post is https://about.fb.com/news/2026/10/introducing-muse-wonder-models/. [Verified · Nov 5, 2026](https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/) - **Usage limits:** Free: 'a usage limit' that refreshes (Meta says 'wait until your free usage limit refreshes'); the amount is not stated. Power plan: 500M Muse tokens per week. Maximum plan: 3B Muse tokens per week. One-time extra tokens 'never expire'. The refresh sentence is from the Muse FAQ on https://ai.meta.com/muse/; the extra-tokens article is https://www.meta.com/help/subscriptions/28153473374286914/ (read 2026-11-05). [Verified · Nov 6, 2026](https://www.meta.com/help/subscriptions/1021145227643680/) - **Data and privacy:** Meta Muse states 'You decide whether we can use your interactions with Muse to develop and improve our AI models. This setting is on when you first use Muse. You can change it at any time using the toggle in your Settings under Data controls.' Meta also states that Muse 'doesn't share your conversations or the data in your virtual machine with Meta ad systems'. Privacy help article shows '8 weeks ago'. The small-business page (read 2026-11-06) also says conversations 'are not shared with Meta's ad systems'. Meta says a Muse Confidential VM is planned 'later this year'. [Provider claim · Nov 5, 2026](https://www.meta.com/help/superintelligence/1047255454427887/) - **Support:** Meta publishes a Muse Help Center and an in-app 'Report an issue' option. No phone line or support hours are stated on the pages read. Help Center is undated. [Verified · Nov 5, 2026](https://www.meta.com/help/superintelligence/1303670544995562/) - Not clear from public sources: Meta's pages differ on status: its small-business page says 'Available today' in the US and Canada, while a help article that showed 'Updated: 8 weeks ago' says Muse subscriptions are 'in limited testing'. The free usage limit is not given as a number, and Meta says Muse 'will have' its own email address without saying when. ### Microsoft Copilot Cowork Microsoft's agent for work in Microsoft 365: it sends email, schedules meetings, builds documents and posts in Teams on your behalf, with scheduled and event-triggered tasks. It is generally available for work or school accounts and in preview for personal Premium and Pro plans in the US and Canada. - Website: https://www.microsoft.com/en-us/copilot/features/cowork - Based: Online - Card on the page: https://allthingscompared.com/ai-agents-and-coworkers/#microsoft-copilot-cowork - Our read (interpretation, not a fact): Microsoft 365 users, including small businesses on Business plans, who want an agent working across Outlook, Teams, OneDrive and SharePoint. - Fits: Open to one person for $50 a month or less; Keeps working after you close the app or laptop; Asks before sensitive actions - **Who can use it today:** Microsoft's Learn overview says 'Cowork for work or school accounts is generally available' and 'Cowork for personal accounts is in preview.' For personal accounts, Microsoft's limits page lists Cowork (in preview) as 'Limited usage' on Microsoft 365 Premium and 'Highest usage' on Pro, 'Not available' on Personal and Family, and says 'Currently only available in the US and Canada.' Work or school accounts need a plan that includes Cowork and an admin who turns it on. The FAQ says 'use and access is currently limited to Anthropic-supported regions' and that education faculty and staff can use it but students cannot. Learn pages show ms.date 2026-11-03 (overview) and 2026-11-04 (FAQ). The limits page (https://support.microsoft.com/en-us/microsoft-365-copilot/ai-credits-and-limits-for-microsoft-365-subscriptions, 'Last updated: October 2026') was read 2026-11-06; the admin requirement is from https://learn.microsoft.com/en-us/copilot/cowork/cowork-admin-governance, read 2026-11-05. The GA blog (June 16, 2026) says 'generally available to Microsoft 365 Copilot customers worldwide today'. [Verified · Nov 6, 2026](https://learn.microsoft.com/en-us/copilot/cowork/) - **Lowest-priced plan that includes it:** Personal accounts (preview): Microsoft 365 Premium, $19.99 per month or $199.99 per year (US dollars), lists Cowork with 'Limited usage'; Microsoft 365 Pro, $99.99 per month ('Monthly billing only'), lists 'Highest usage'. Work or school accounts: Microsoft 365 Business Standard with Copilot, $23.50 per user per month paid yearly ($28.20 paid monthly), includes access to Cowork. Cowork is also metered: Microsoft's FAQ says it 'uses a usage-based billing model' and the June 2026 blog says 'PayGo is priced at $0.01 per Copilot Credit'. No per-task price is published. The individuals pricing page names 'agentic AI' under Premium and Pro and does not name Cowork; the plan-by-plan Cowork row is on the limits page. The Business Standard figures are from https://www.microsoft.com/en-us/copilot/pricing/business (read 2026-11-05). A Copilot Business add-on shows a first-year price of $18.00 per user per month until December 31, 2026 (regular $21.00), per that page. [Verified · Nov 6, 2026](https://www.microsoft.com/en-us/copilot/pricing/individuals) - **What it does:** Microsoft's Learn overview says Cowork 'sends emails, schedules meetings, creates documents, posts in Teams, and manages your calendar'. It also searches your organization, manages OneDrive and SharePoint files, does deep research, prepares daily briefings, drafts stakeholder communications and 'runs prompts on a schedule'. It can also complete web tasks in Microsoft Edge on your device. Product page examples: prep a customer meeting, triage your inbox, coordinate a launch. [Verified · Nov 6, 2026](https://learn.microsoft.com/en-us/copilot/cowork/) - **Background and scheduled work:** Microsoft Copilot Cowork states that 'your tasks keep running even when your laptop is off' (Microsoft 365 blog, June 16, 2026). Microsoft's FAQ lists its speed as 'Minutes to hours (autonomous execution)' and describes scheduled prompts ('Send me a daily briefing every morning at 9 AM') and event-driven tasks that run when a matching email arrives or a Teams message is posted. 'By default, event-driven tasks prepare actions for your approval rather than acting on their own.' The FAQ is https://learn.microsoft.com/en-us/copilot/cowork/cowork-faq (ms.date 2026-11-04). [Provider claim · Nov 6, 2026](https://www.microsoft.com/en-us/microsoft-365/blog/2026/06/16/copilot-cowork-is-now-generally-available/) - **Apps and connectors:** Microsoft's FAQ lists built-in skills for Word, Excel, PowerPoint, PDF, Email, Scheduling, Calendar Management, Meetings, Daily Briefing, Enterprise Search and Deep Research, and works with Teams, OneDrive and SharePoint. Plugins from the Microsoft 365 App Store 'add skills, connectors, or both', and Microsoft's plugin catalog lists partner plugins including Slack, Gmail, Google Drive, Notion, Salesforce and Zoom. Partner plugins 'require your IT admin to make them available'. No count of plugins is stated. The partner list is from https://learn.microsoft.com/en-us/copilot/cowork/cowork-available-plugins (ms.date 2026-11-02, read 2026-11-05), which says the catalog 'continues to grow'. [Verified · Nov 6, 2026](https://learn.microsoft.com/en-us/copilot/cowork/cowork-faq) - **Own identity and access:** Microsoft's FAQ says 'Cowork inherits your permissions, so it can access only the files and emails you can already access.' Its admin page says each automated task runs as the user who created it. No own account, email address or calendar for Cowork is described on the pages read. Microsoft's separate Autopilot (formerly Scout) is described with 'its own identity' but is in private preview (set aside). [Verified · Nov 6, 2026](https://learn.microsoft.com/en-us/copilot/cowork/cowork-faq) - **Approvals and controls:** Microsoft Copilot Cowork states 'Before Cowork takes a sensitive action like sending an email or posting in Teams, it displays an approval prompt', with a risk level indicator for medium and high risk actions. You can approve once, skip similar prompts for the rest of the session, or cancel, and you can pause, resume or cancel a task. Admins can disable Cowork per user, set spending policies and turn off individual models. Statements are Microsoft's own and were not tested. The overview adds 'You approve each action before it happens.' [Provider claim · Nov 6, 2026](https://learn.microsoft.com/en-us/copilot/cowork/cowork-faq) - **Models named:** Microsoft's FAQ says the model picker defaults to 'Auto' and 'The choices include GPT 5.5, GPT 5.6, and GPT 6 variants, Claude Opus, Claude Sonnet, Claude Fable, and Grok 4.7'; inside the EU Data Boundary 'only GPT models are supported'. Image generation uses ChatGPT Images 2.0. The June 16, 2026 blog said Cowork then ran on Anthropic models. Model availability varies by tenant configuration. [Verified · Nov 6, 2026](https://learn.microsoft.com/en-us/copilot/cowork/cowork-faq) - **Usage limits:** For work accounts, Microsoft's FAQ says activities such as 'model responses, tool and skill calls, image generation, and browser tasks count toward consumption' and that admins can set limits per user or group. For personal accounts Premium has 'Limited usage' and Pro 'Highest usage', with no numbers. Limits by design: attached files under 200 MB, prompts up to 250,000 characters, and Cowork cannot edit local files or delete OneDrive or SharePoint files. Personal-account wording is from the limits page (Last updated: October 2026). [Verified · Nov 6, 2026](https://learn.microsoft.com/en-us/copilot/cowork/cowork-faq) - **Data and privacy:** Microsoft Copilot Cowork states that its data 'stays in your tenant and respects existing user and admin permissions' (product page FAQ) and, on the Business pricing page, 'Prompts and responses are not used to train the underlying large language models'. Microsoft's FAQ says Cowork can use Anthropic Claude models 'as a subprocessor', that 'Some models, such as Claude Fable, require data retention', and that use is limited to Anthropic-supported regions. The product-page and pricing-page statements were read 2026-11-05. For personal accounts, Microsoft's pages defer to its Privacy Statement, which was not read. [Provider claim · Nov 6, 2026](https://learn.microsoft.com/en-us/copilot/cowork/cowork-faq) - **Support:** Microsoft's Business pricing page lists a sales and pricing contact by chat and phone ('Available M-F 6 AM to 6 PM PT'). The Cowork pages read state no support channels or hours for the product itself; Learn describes in-product feedback with thumbs up or down and comments. The hours apply to the sales line, not to product support. [Verified · Nov 5, 2026](https://www.microsoft.com/en-us/copilot/pricing/business) - Not clear from public sources: Microsoft's pages publish no price per task (usage is billed in Copilot Credits at $0.01 each) and give no count of plugins. The work route needs an admin to turn Cowork on and set up usage-based billing. The June 2026 blog says 'worldwide' while the FAQ limits use to Anthropic-supported regions. ### Mistral Vibe Work Work mode of Mistral's Vibe (formerly Le Chat): a web and mobile agent for multi-step tasks across connected apps, with scheduled tasks in Public Preview. Free, Pro ($14.99 a month) and Team ($24.99 per user a month) plans. - Website: https://mistral.ai/products/vibe - Based: Online - Card on the page: https://allthingscompared.com/ai-agents-and-coworkers/#mistral-vibe-work - Our read (interpretation, not a fact): People who want a work agent on a low monthly price, with connectors and MCP support, in a single Mistral account. - Fits: Open to one person for $50 a month or less; Asks before sensitive actions - **Who can use it today:** Mistral's pricing page lists Free, Pro and Team plans, with Enterprise by contact, and its product page FAQ says Free covers 'quick answers' and 'simple tasks' while upgrading to Pro unlocks agent features and extended multi-step work. Mistral's docs describe Work as 'Vibe's productivity mode for delegating complex, multi-step tasks across your apps and tools' and label scheduled tasks 'Public Preview'. The pricing table labels connectors 'beta'. No waitlist wording or region limit appears on the pages read. Mistral's help article (https://help.mistral.ai/en/articles/682992-le-chat-is-now-vibe) says 'Le Chat is now Vibe'. The pricing page showed a France country selector with USD prices, so it is not a verified en-US view; the same list prices appear in the May 28, 2026 launch post. Mistral's pages say Work is on web and mobile; chat.mistral.ai returned 403 to plain fetches on 2026-11-01 and was not opened. [Verified · Nov 6, 2026](https://mistral.ai/products/vibe) - **Lowest-priced plan that includes it:** Pro, $14.99 per month (US dollars as displayed, excluding taxes; the pricing page does not say which billing it shows). Team is $24.99 per user per month. Free is $0 with limited use; verified students pay $5.99 per month for up to 12 months. Mistral's product page says agent features and long-running tasks start at Pro, while the pricing table lists some scheduling and connectors on Free. Mistral's help centre says annual billing is 20% cheaper than monthly for Pro and Team (https://help.mistral.ai/en/articles/455205-how-can-i-upgrade-or-cancel-my-subscription, read 2026-11-01), so the displayed price may be the monthly one; Mistral does not say. The student price is from the pricing FAQ read 2026-11-05. [Verified · Nov 6, 2026](https://mistral.ai/pricing) - **What it does:** Mistral's product page says 'Schedule recurring workflows and tasks. Package workflows into reusable agents, or set tasks to run on a trigger.' Mistral's docs give example prompts: 'Summarize this 80-page contract and flag the renewal risks', 'Draft replies to my unread emails from the legal team' and 'Find the latest customer feedback in Notion, group it by theme, and prepare a slide deck'. The example prompts are from https://docs.mistral.ai/vibe/work/get-started, read 2026-11-05. Vibe also has a Code mode for developers, which is set aside as a coding agent. [Verified · Nov 6, 2026](https://mistral.ai/products/vibe) - **Background and scheduled work:** Mistral Vibe Work states 'Scheduled tasks let you run a prompt automatically at a future date or on a recurring schedule' and 'A scheduled task runs unattended, so think about approvals upfront.' Scheduled tasks are labelled 'Public Preview', and the docs say how many you can have 'depends on your plan'. Mistral's pages do not say whether the computer must be on, or whether an ordinary Work task keeps running after you close the app. Triggers are time-based for now, with event-based triggers possible later, per the same docs page read 2026-11-05. [Provider claim · Nov 6, 2026](https://docs.mistral.ai/vibe/work/scheduled-tasks) - **Apps and connectors:** Mistral Vibe Work states 'Vibe integrates with 100+ tools including email, calendar, Slack, GitHub, Jira, and more through connectors and full MCP compatibility.' Mistral's docs name Gmail, Outlook, Slack, Notion, Linear, GitHub, Atlassian, Google Drive, SharePoint and Stripe; the pricing table lists connectors (beta) and custom MCP connectors (beta), and a connector directory exists. The count is Mistral's own figure. The named list is from https://docs.mistral.ai/vibe/work/connectors (read 2026-11-05). [Provider claim · Nov 6, 2026](https://mistral.ai/products/vibe) - **Own identity and access:** Mistral's docs say Work connects to apps through accounts you authorize with OAuth ('Your password is never shared with us'). No own account, email address or calendar for Vibe Work is described on the pages read. Workspace admins can disable specific connectors for the whole workspace. [Verified · Nov 5, 2026](https://docs.mistral.ai/vibe/work/connectors) - **Approvals and controls:** Mistral Vibe Work states 'Before Work performs an action that touches an external system, it stops and asks for your approval', covering anything that 'creates, modifies, sends, posts, or deletes'. You can choose Continue (this action only), Always allow (per function) or Decline. The docs add 'There is no upfront plan to approve: Work starts immediately, and you watch the steps appear'. Workspace admins can disable specific connectors. The launch post (https://mistral.ai/news/vibe-agent, May 28, 2026) says Work 'maps out a plan and gets your sign-off before it starts'; the docs page says otherwise. Statements are Mistral's own and were not tested. [Provider claim · Nov 6, 2026](https://docs.mistral.ai/vibe/work/safety-and-approvals) - **Models named:** Mistral's launch post says 'Vibe runs on flagship Mistral models optimized for reasoning, agentic tasks, tool calls, and coding', and its pricing FAQ says 'For most tasks and coding: Mistral Medium.' Which model a Work task uses is not stated. Launch post dated May 28, 2026. [Verified · Nov 5, 2026](https://mistral.ai/news/vibe-agent) - **Usage limits:** Mistral's pricing page says plans are 'Subject to fair usage limits and Mistral's Terms of Service.' Free lists 'Limited messages and web searches'; Pro lists messages 'Up to 6x free' and web searches 'Up to 5x free'; task scheduling is shown as 'Up to 5' and 'Unlimited' without a clear plan mapping in the text read. Paid plans can 'Extend usage with PAYG credits'. The pricing FAQ gives one example: 'Flash answers: 200 / day on Team vs. 150 / day on Pro' (read 2026-11-05). [Verified · Nov 6, 2026](https://mistral.ai/pricing) - **Data and privacy:** Mistral Vibe Work states 'Vibe: users are not opted out by default and can opt out at any time in their settings' for training on input and output data, and 'Vibe (Enterprise): customers are opted out of training by default'. The pricing page lists 'Model training: Opt-out'. Temporary Chat 'isn't saved to your history or used to train our models', and zero data retention 'is not available for Vibe Work or Chat'. Help article read 2026-11-05; the pricing page row was read 2026-11-06. The zero-data-retention sentence is from https://help.mistral.ai/en/articles/347612-can-i-activate-zero-data-retention-zdr. [Provider claim · Nov 5, 2026](https://help.mistral.ai/en/articles/455207-can-i-opt-out-of-my-input-or-output-data-being-used-for-training) - **Support:** Mistral's help centre says the fastest route is the help widget and 'Support is only available to users with an active account'; response times 'may vary'. The pricing page lists chat and email support for Pro and Team. A Discord community is named. Support hours are not stated. [Verified · Nov 5, 2026](https://help.mistral.ai/en/articles/347458-how-do-i-contact-support) - Not clear from public sources: Mistral's pages do not say whether an ordinary Work task keeps running after you close the app, and they give scheduled-task limits only as 'Up to 5' and 'Unlimited' without a clear plan mapping. The pricing page showed a France country selector with US-dollar prices. The launch post says Work gets your sign-off before it starts; the docs say there is no upfront plan to approve. ### Perplexity Computer Perplexity's agent that researches, builds documents and websites, manages email and runs scheduled tasks in the cloud, billed in credits. Personal Computer (a Mac and Windows app, with a Mac mini always-on option) and Portable Computer (local, on NVIDIA hardware) are versions of it on the same plans. - Website: https://www.perplexity.ai/help-center/en/articles/13837784-what-is-computer - Based: Online - Card on the page: https://allthingscompared.com/ai-agents-and-coworkers/#perplexity-computer - Our read (interpretation, not a fact): People on a Pro or Max plan who want one agent across the web, connected apps and their own desktop. - Fits: Open to one person for $50 a month or less; Keeps working after you close the app or laptop; Asks before sensitive actions - **Who can use it today:** Perplexity's help article says 'Computer is available on paid plans, with no setup or installation needed', on web desktop, iOS and Android, and that 'Some free accounts also include a limited number of free Computer turns', a one-time allowance. Personal Computer, a desktop app for Mac and Windows (on a Mac mini it 'runs 24/7 in the background'), and Portable Computer, which runs locally on NVIDIA hardware, are included with the same plans. An April 16, 2026 Perplexity post said Personal Computer was 'rolling out today to Perplexity Max subscribers' with a waitlist; the current Mac page says 'Available now.' Perplexity's main site returned 403 to plain fetches on 2026-11-01; help-centre pages were read through a fetch tool and the browser pane. The Personal Computer and Portable Computer wording is from https://www.perplexity.ai/help-center/en/articles/14659663-what-is-personal-computer and https://www.perplexity.ai/help-center/en/articles/20260915-what-is-portable-computer, read 2026-11-05. [Verified · Nov 6, 2026](https://www.perplexity.ai/help-center/en/articles/13837784-what-is-computer) - **Lowest-priced plan that includes it:** Pro, $17 per month when billed annually (US dollars; a month-to-month Pro price is not shown on the pages read). Max is $167 per month billed annually or $200 billed monthly. Credits cost 100 per $1: Pro has no monthly credit allotment ('paid plans include weekly Computer usage', rolling out gradually) and a one-time 4,000 bonus credits; Max includes 10,000 credits per month and a one-time 35,000 bonus. A quick fact costs about 15 to 70 credits and a multi-step research report about 350 to 950. Pricing page undated. The credit figures were read on https://www.perplexity.ai/help-center/en/articles/13838041-how-credits-work-on-perplexity and the Max price on https://www.perplexity.ai/help-center/en/articles/11680686-perplexity-max, both 2026-11-06; the credits article says the task costs are 'examples, not guarantees'. Enterprise Pro is $34 per seat per month billed annually. The pro page showed limited-time bonus Computer credits as a promotion. [Verified · Nov 5, 2026](https://www.perplexity.ai/hub/pricing) - **What it does:** Perplexity's help article says 'Computer is an independent digital worker that completes tasks and workflows for you. Instead of just providing information, it takes action.' Perplexity's pages describe researching, building documents, apps and websites, managing email and running tasks on a schedule; an example prompt is 'Research our top 5 competitors, compare their pricing, create a slide deck with the findings, and email it to the team.' The example prompt is from the same help article as read 2026-11-05. [Verified · Nov 6, 2026](https://www.perplexity.ai/help-center/en/articles/13837784-what-is-computer) - **Background and scheduled work:** Perplexity Computer states 'Computer runs tasks in the background while you're away', and of Automations that 'Computer runs each task autonomously in the cloud, so your laptop doesn't need to be open', as often as every 15 minutes. If credits run out, active tasks pause. Personal Computer on a Mac mini 'runs 24/7 in the background, working even when your laptop is closed' (the Mac mini must stay on). Automations replace the older Perplexity Tasks. The first sentence is on the 'What is Computer' article, read 2026-11-06. [Provider claim · Nov 6, 2026](https://www.perplexity.ai/help-center/en/articles/11521526-perplexity-tasks) - **Apps and connectors:** Perplexity Computer states that it works 'through hundreds of app connectors' (help article), and its pricing page lists 'Connect 400+ app connectors'. The help article names Gmail, Outlook, GitHub, Linear, Slack and Notion; other Perplexity pages name Snowflake, Databricks and Salesforce, side panels in Excel, Word, PowerPoint and Outlook, Slack, and a Plugins directory. The counts are Perplexity's own figures. The '400+' line is from https://www.perplexity.ai/hub/pricing and the other names from Perplexity's Computer blog and help pages, read 2026-11-05. [Provider claim · Nov 6, 2026](https://www.perplexity.ai/help-center/en/articles/13837784-what-is-computer) - **Own identity and access:** Perplexity's pages say Computer in Email 'runs tasks for verified senders under their existing Perplexity accounts'; the shared address computer@perplexity.com is Perplexity's inbound address for sending tasks and it replies only to the verified sender. Computer runs in an isolated cloud sandbox with a dedicated filesystem and browser (Enterprise article). No own email address, account or calendar for the agent is described. Perplexity's separate Teammates product has its own desktop and browser; see that option. [Verified · Nov 5, 2026](https://www.perplexity.ai/help-center/en/articles/13901210-computer-for-enterprise) - **Approvals and controls:** Perplexity Computer states 'Perplexity asks you to approve consequential writes before they run.' When you create an automation, Computer shows the schedule and what it will do and you 'Approve to create the task.' Perplexity says Personal Computer for Windows 'always alerts a user before it takes a sensitive action like sending an email or deleting a file'. Enterprise admins can turn Computer off, control connectors and see activity in audit logs. The troubleshooting article is the page that states the consequential-writes approval. The Windows sentence is from https://www.perplexity.ai/hub/blog/personal-computer-on-windows (July 28, 2026) and the admin controls from the Enterprise article, read 2026-11-05. Statements are Perplexity's own. [Provider claim · Nov 6, 2026](https://www.perplexity.ai/help-center/en/articles/2026091802-troubleshooting-connectors) - **Models named:** Perplexity says Computer 'creates teams of agents across 20+ frontier models'. Its pricing page names models including GPT-6.1 Sol, Gemini 3.8 Flash, Claude Sonnet 5.5 and Kimi K3. A changelog entry (August 24, 2026) names GPT-5.6 Terra as the default for Computer subagents and GPT-5.6 Luna for scheduled automations. Local modes use PPLX 27B, Qwen and Gemma models. Model names differ between Perplexity pages and were not reconciled. [Verified · Nov 5, 2026](https://www.perplexity.ai/personal-computer) - **Usage limits:** Perplexity's credits article says '100 credits equals $1' and that on Pro 'Included usage is rolling out gradually, so some accounts may not have it yet'. Max has a monthly spending cap on credit charges; 'The default is $200, and you can set it anywhere up to $5,000'. Free turns are a one-time allowance. Out-of-credit accounts have active tasks paused and new tasks blocked. Effort settings run from Light to Ultra. The effort settings are from the 'What is Computer' collection, read 2026-11-05. [Verified · Nov 6, 2026](https://www.perplexity.ai/help-center/en/articles/13838041-how-credits-work-on-perplexity) - **Data and privacy:** Perplexity Computer states for Enterprise 'Your organization's data is never used to train Perplexity's models. This guarantee extends fully to Computer'. For individual plans, Perplexity's plan article lists 'Opt-out AI training available' for Pro and Max. It also states that Computer 'executes actions in an isolated sandbox environment'. No Computer-specific retention statement for individual plans was found. Plan article 'Last modified' November 5, 2026 when read. The Enterprise guarantee is from the Enterprise article. [Provider claim · Nov 5, 2026](https://www.perplexity.ai/help-center/en/articles/11187416-which-perplexity-subscription-plan-is-right-for-you) - **Support:** Perplexity says to contact support through Help > Contact Support and include the task link. Its plan article lists support as Basic for Free, Priority for Pro and Dedicated for Max, and says Pro subscribers can use live chat in Settings. Hours are not stated. Plan article read 2026-11-05. [Verified · Nov 5, 2026](https://www.perplexity.ai/help-center/en/articles/11187416-which-perplexity-subscription-plan-is-right-for-you) - Not clear from public sources: Perplexity's pages show no month-to-month Pro price, only $17 a month billed annually. Pro has no monthly credit allotment, and its weekly included usage is 'rolling out gradually'; heavy use is bought as credits. Model names differ between Perplexity pages, and no Computer-specific retention wording for individual plans was found. ### Perplexity Teammates Perplexity's named, always-on agents that each own an ongoing job such as a campaign, bug triage, metrics tracking or market research, each with its own desktop and browser. A web-only research preview for individual Max subscribers. - Website: https://www.perplexity.ai/hub/products/teammate - Based: Online - Card on the page: https://allthingscompared.com/ai-agents-and-coworkers/#perplexity-teammates - Our read (interpretation, not a fact): Individual Max subscribers who want agents that own recurring responsibilities rather than single tasks. - Fits: Has its own computer or email address - **Who can use it today:** Perplexity's product page says 'Teammates are available today for Perplexity Max subscribers, on the web. Coming to iOS and Android.' Its help article says the preview is limited to individual Max subscribers on the web ('The initial research preview requires an individual Perplexity Max subscription') and that 'Enterprise Pro and Enterprise Max are not included by default in this launch.' Regions are not stated. The product page is undated. The help article (https://www.perplexity.ai/help-center/en/articles/2026102601-perplexity-teammates) was read 2026-11-05 and 2026-11-06 (it showed 'Last modified' November 5, 2026); it calls the feature a research preview that is 'still under development', with behavior that 'may change'. No waitlist wording was found. [Verified · Nov 6, 2026](https://www.perplexity.ai/hub/products/teammate) - **Lowest-priced plan that includes it:** Max only: $167 per month when billed annually, or $200 per month billed monthly (US dollars). Perplexity states 'For a promotional period, every Max plan will include free usage for Perplexity teammates'; afterwards teammates 'use the credits that are already included with your Max plan' (10,000 credits per month on Max; 100 credits equals $1). There is no separate teammate price. The length of the promotional period is not stated. The $200 monthly price and the 10,000-credit figure are from https://www.perplexity.ai/help-center/en/articles/11680686-perplexity-max and the credits article, read 2026-11-06; the annual figure is from the pricing page, read 2026-11-05. [Verified · Nov 6, 2026](https://www.perplexity.ai/hub/products/teammate) - **What it does:** Perplexity's product page says 'Teammates are a team of AI agents that own ongoing work', for example 'managing a growth campaign, triaging incoming bug reports, tracking your metrics, or researching a new market'. Each gets 'a role, a standing brief, and rules for when to bring you in'. Examples on the page: an on-call engineer that investigates alerts and opens a fix pull request, an ads manager, and a legal research assistant. The page lists coding projects as one use among ad campaigns, research, sales and operations. [Verified · Nov 6, 2026](https://www.perplexity.ai/hub/products/teammate) - **Background and scheduled work:** Perplexity Teammates states 'Teammates work 24/7 in the background and notify you when decisions need to be made', and that they are 'Proactive by design, they run on automations and keep moving work forward on your behalf'. Each teammate keeps 'a single thread that never closes'. The pages read do not say whether your own computer must stay on. Because the pages do not say the work continues with your device closed, this option is not tagged for that need. [Provider claim · Nov 6, 2026](https://www.perplexity.ai/hub/products/teammate) - **Apps and connectors:** Perplexity's product page says 'Teammates use the same skills and plugins you have installed, with the same permissions' (Computer's plugins and connectors), and that teammates use 'full desktop and browser use'. The help article says teammates can be tagged in Slack, GitHub and Linear. The pages read give no separate app count for teammates. Computer's pricing page lists 'Connect 400+ app connectors'; see the Perplexity Computer option. [Verified · Nov 6, 2026](https://www.perplexity.ai/hub/products/teammate) - **Own identity and access:** Perplexity's product page says 'Teammates each own their own desktop and browser', and its help article says 'Each teammate runs on its own isolated computer, with its own desktop and browser'. Each teammate has its own role. The pages state no own email address, account or calendar; teammates use your installed plugins 'with the same permissions'. Browsing sessions and cookies are independent for each teammate. [Verified · Nov 6, 2026](https://www.perplexity.ai/hub/products/teammate) - **Approvals and controls:** Perplexity Teammates states that teammates 'notify you when decisions need to be made' and 'reach out when it needs your input', and that you 'can take over and steer its work'; when connecting a tool you are told to 'Review the requested permissions before approving access'. The pages read do not state a general approval step before sensitive actions; the product page's sample shows a fix that 'needs your approval before it ships'. Perplexity's Computer help says it asks you to approve consequential writes; Perplexity says teammates are 'built on the same security and privacy infrastructure as Computer', but no teammate-specific approval rule was found, so this option is not tagged for asking before sensitive actions. [Provider claim · Nov 6, 2026](https://www.perplexity.ai/help-center/en/articles/2026102601-perplexity-teammates) - **Models named:** Perplexity's help article says 'New teammates use Claude Opus 5.5 by default.' You can choose a different model for each teammate, and agents a teammate starts use that teammate's model unless you say otherwise. The product page's sample screen shows GPT-6 Astra. Model names change often. [Verified · Nov 6, 2026](https://www.perplexity.ai/help-center/en/articles/2026102601-perplexity-teammates) - **Usage limits:** Perplexity's help article says 'Max subscriptions include a monthly credit allocation that teammates can use', and that complex projects can use more credits, 'particularly when agents run in parallel'. No usage numbers specific to teammates are stated; Max includes 10,000 credits per month. A Max monthly spending cap on credit charges applies (default $200). [Verified · Nov 6, 2026](https://www.perplexity.ai/help-center/en/articles/2026102601-perplexity-teammates) - **Data and privacy:** Perplexity Teammates states that teammates are 'Built on the same security and privacy infrastructure as Computer'. For individual plans, Perplexity's plan article lists 'Opt-out AI training available' for Max. No teammate-specific retention wording was found. The plan-article line was read 2026-11-05. [Provider claim · Nov 6, 2026](https://www.perplexity.ai/hub/products/teammate) - **Support:** Perplexity's plan article lists Priority support for Pro and Dedicated support for Max, with 'dedicated infrastructure and faster response times' described for Max; contact is through Help > Contact Support. Hours are not stated. No teammate-specific support terms were found. [Verified · Nov 5, 2026](https://www.perplexity.ai/help-center/en/articles/11680686-perplexity-max) - Not clear from public sources: Perplexity's pages say 'available today' for Max subscribers on the web and call it a research preview, and one setup line says 'with access to the research preview'. They give no region list, no usage numbers specific to teammates, no general approval step before sensitive actions, and no own email address or calendar. ### Tasks in Gemini Google's task agent in the Gemini app, named Gemini Spark during its beta: hand it a goal and it plans, browses, and runs scheduled or monitoring tasks. It needs a personal Google Account on Google AI Pro or Ultra. - Website: https://support.google.com/gemini/answer/17094507 - Based: Online - Card on the page: https://allthingscompared.com/ai-agents-and-coworkers/#tasks-in-gemini - Our read (interpretation, not a fact): Google AI Pro or Ultra subscribers on a personal account who want an agent for inbox, shopping, booking and monitoring tasks. - Fits: Open to one person for $50 a month or less; Asks before sensitive actions - **Who can use it today:** Google's help article says you need 'a Google AI Pro or Ultra subscription' and a personal Google Account ('this feature isn't available if you sign in with a work or school Google Account'), to be 18 or over and to keep Activity on. It is available wherever Gemini Apps are supported 'except in the European Economic Area, Nigeria, Switzerland, and the United Kingdom', and 'Gemini's ability to handle tasks is experimental.' Google's help article says 'Since May, we've been testing Spark in beta with our subscribers' and 'we renamed it Task'. Google's plans page lists Tasks in Gemini only under Ultra ('first access to advanced features like Tasks in Gemini (in select countries)'). The help page is undated (copyright 2026 Google). Google's Google One page (https://one.google.com/about/google-ai-plans/, read 2026-11-05) says 'Available to Google AI Pro and Ultra subscribers in select countries'. Google's May 19, 2026 post said the Spark beta would roll out to US Google AI Ultra subscribers first. [Verified · Nov 6, 2026](https://support.google.com/gemini/answer/17094507) - **Lowest-priced plan that includes it:** Google AI Pro, $19.99 per month (US dollars, monthly; the Google One page also shows $199.99 per year). Google AI Ultra starts at $99.99 per month ($199.99 for the 20x tier), billed monthly only. The help article says Pro or Ultra is needed; Google's plans page lists Tasks in Gemini under Ultra only, and its Google One page lists 'Access to Tasks in Gemini' under Pro. Both pricing pages are undated. Google AI Plus ($4.99 per month) is not listed as including Tasks. [Verified · Nov 6, 2026](https://gemini.google/subscriptions/) - **What it does:** Google's help article says 'Assign Gemini tasks to automate complex workflows and manage schedules for ongoing tasks.' Examples: declutter your inbox by summarizing or archiving newsletters, make a to-do list after a vacation, unsubscribe from email lists, alert you when flight and hotel prices drop within a budget, and tasks on websites such as shopping, booking travel or making restaurant reservations. In the Gemini Mac app it can manage files on your Mac. The Mac wording is from https://support.google.com/gemini/answer/17208717, read 2026-11-05. [Verified · Nov 6, 2026](https://support.google.com/gemini/answer/17094507) - **Background and scheduled work:** Tasks in Gemini states 'A schedule lets you hand over tasks to Gemini so they run automatically in the background' (time-based schedules, Gmail monitors and topic monitors). Google's help article says 'Gemini continues to work on your task even if you close your device' for tasks on a remote browser (used on mobile), but tasks in a local Chrome browser need the device and Chrome awake and running, and 'Schedules will not run as long as task mode is off or if your device is off.' Up to 15 tasks can run at a time. The first quoted sentence and the schedule types are from https://support.google.com/gemini/answer/17094710 (read 2026-11-05). Because the same page says schedules do not run if the device is off, this option is not tagged for keeping going when closed. [Provider claim · Nov 6, 2026](https://support.google.com/gemini/answer/17094507) - **Apps and connectors:** Google's help article lists Contacts, Gemini Notebook, Google Photos, 'Google Workspace, which includes Calendar, Docs, Drive, Gmail, Keep, Sheets, Slides, and Tasks', Google Search services (Search, Finance, Flights, Hotels and Maps), YouTube, custom apps and 'Several third-party apps'. Google's June 30, 2026 post names Canva, Dropbox, Instacart, OpenTable and Zillow Rentals and support for custom MCP. No total number of apps is stated. The June 30 post is https://blog.google/innovation-and-ai/products/gemini-app/gemini-spark-updates-june-2026/, read 2026-11-05. [Verified · Nov 6, 2026](https://support.google.com/gemini/answer/17094507) - **Own identity and access:** Google's help article says 'Gemini uses your info from sources like Connected Apps, skills, chats, tasks, websites you're logged into, Personal Intelligence, location, and more.' Tasks act as you. Google's May 2026 post listed 'texting and emailing Spark' as planned abilities; the pages read describe no own account, email address or calendar for the agent today. Google's enterprise Gemini agent with coworker agents that get their own Workspace accounts is not publicly available and is set aside. [Verified · Nov 6, 2026](https://support.google.com/gemini/answer/17094507) - **Approvals and controls:** Tasks in Gemini states 'Gemini is designed to always ask for your review and confirmation before it completes certain actions', listing sending communications, modifying or deleting your data, making purchases, completing financial transactions and submitting web forms. For tasks that involve web browsing, 'Gemini will ask for confirmation first'. Permission choices are don't allow, allow once, allow for this chat, and always allow. The same page says 'Gemini can perform bulk actions on private tasks in Google Tasks without your confirmation' and that if a schedule runs while you are offline, 'you may not be able to stop Gemini from completing an unintended action'. Statements are Google's own and were not tested. Permission choices are from https://support.google.com/gemini/answer/18766567, read 2026-11-05. There are no admin controls because work and school accounts are excluded. [Provider claim · Nov 6, 2026](https://support.google.com/gemini/answer/17094507) - **Models named:** Google's May 19, 2026 post said Gemini Spark 'runs on Gemini 3.5 and uses the Antigravity harness'. The help article lets you change 'the Gemini model or thinking level'. Which model Tasks in Gemini uses today is not stated. The plans page names Gemini 4 Argon and Flash models for the Pro plan without tying them to Tasks. [Verified · Nov 5, 2026](https://blog.google/innovation-and-ai/products/gemini-app/next-evolution-gemini-app/) - **Usage limits:** Google's help article says 'You can have up to 15 tasks running at a given time.' Gemini Apps have 'compute-based usage limits' that rise with higher Google AI plans: Pro is 4x higher usage than Free, and Ultra is 5x or 20x higher than Pro. Limits refresh every 5 hours until a weekly limit, and can be extended with AI credits. The usage multiples are from https://gemini.google/subscriptions/ (read 2026-11-06); no task or message counts are given. [Verified · Nov 6, 2026](https://support.google.com/gemini/answer/17094507) - **Data and privacy:** Tasks in Gemini states that you must keep Activity on to use it. Google's Gemini Apps Privacy Hub says 'A subset of chats are reviewed by human reviewers (including Google's trained service providers) to help improve Google services', that 'Reviewed chats are retained for up to three years', and that auto-delete defaults to 18 months. The help article says that when acting for you 'Gemini will share necessary info with third parties'. The Keep Activity requirement is on the Tasks help page (read 2026-11-06). The Privacy Hub is undated. [Provider claim · Nov 5, 2026](https://support.google.com/gemini/answer/13594961) - **Support:** Google's help article links the Gemini Apps Help Center, a help community and a feedback link. No support hours or phone line are stated. No Tasks-specific support terms were found. [Verified · Nov 6, 2026](https://support.google.com/gemini/answer/17094507) - Not clear from public sources: Google's plan pages differ on whether Pro includes it: the Gemini plans page lists it under Ultra, while the help article and the Google One page include Pro. Google calls the feature experimental, excludes work or school accounts, and says schedules do not run if the device is off. ## Questions ### How were these agents chosen? Each is an agent product from a maker of one of the assistants on the AI chat assistants page (Anthropic, Brave, Google, Mistral, OpenAI, Perplexity, Proton and SpaceXSI, the company behind Grok) or from Microsoft or Meta. To be considered, the maker's own pages had to say the product carries out multi-step tasks for a user or team on its own, let a person or business start on the check date by paying the published price or on a free tier without a sales call, waitlist or private-preview invitation, and state its price or access terms. A public preview or beta that the pages say holders of a named plan can use today counts as available, and the option quotes the status wording. A page that says 'rolling out' but also says holders of a named plan can use the product now counts the same way for every product (Dots in ChatGPT, Claude Cowork and Meta Muse each use that wording). More than twenty products met those tests, so one extra rule was applied to every candidate: the product is a ready-made agent for general work that one person can start alone, meaning the maker's pages present it for work across the person's apps and files (not only code, not only a web browser, not a tool for building agents, not an operations platform for one cloud) and list a plan or free tier that one person can buy, with no minimum number of seats. Ten products meet it; the page lists them in alphabetical order and does not rank them. Set aside by that rule, and researched in full in the research notes so they can be swapped in: coding agents (Claude Code, OpenAI Codex, GitHub Copilot cloud agent, Meta Muse Code, Mistral Vibe Code, Grok Build, Cursor Cloud Agents, Google Antigravity); browser agents (Claude in Chrome, Chrome auto browse, Perplexity Comet, Browse with Copilot, Brave AI browsing); team-only agents (Claude Tag, OpenAI workspace agents, the agents in Gemini Enterprise Business edition); Microsoft's Azure SRE Agent; report and digest tools (Microsoft Researcher and Analyst, Google Daily Brief); and tools for building agents (Google Workspace Studio, Microsoft Copilot Studio, Claude Managed Agents, Mistral Studio, the OpenAI Agents SDK). Set aside because the maker's pages show no public availability on the check date, with the maker's words: Claude Duet from Anthropic ('Claude Duet is a research preview on a waitlist for Claude Pro, Max, and Ultra subscribers in the US.', read 2026-11-05); the Gemini agent with coworker agents from Google Cloud ('While these new Gemini agent capabilities are not publicly available yet, startups can contact our sales team', October 15, 2026); Microsoft Autopilot, formerly Scout ('Available in Private Preview.', read 2026-11-05) and Copilot Code ('Code is rolling out to Frontier at the end of the month', September 25, 2026); OpenAI specialist dots ('We're starting with focused enterprise pilots.', September 29, 2026) and Daybreak ('Daybreak access requires Trusted Access for Cyber approval.', read 2026-11-05); Meta's Muse on AI glasses ('In the coming months we're bringing Muse to our AI glasses'), the Meta Business Agent briefing for owners ('You can join the waitlist here', June 3, 2026) and the agent features in the Meta AI app ('start rolling out today in select markets', July 24, 2026); and Claude Security ('Available in public beta for Claude Enterprise', with a 'Contact sales' button, read 2026-11-05). Ended or ending: Google Jules ('Jules will wind down on December 31, 2026', October 29, 2026), OpenAI ChatGPT agent ('ChatGPT agent is no longer available.', read 2026-11-05), Operator, Atlas and Pulse, Gemini CLI for individual plans, Google Project Mariner (no current page), Perplexity Email Assistant ('Email Assistant will soon be deprecated', read 2026-11-06) and Perplexity Labs (its only availability line is dated May 29, 2025). Perplexity Personal Computer and Portable Computer are versions of Perplexity Computer on the same plans, so they are covered in that one option. Proton lists no agent product, and its Custom Lumos are saved-instruction assistants. ### Where do the prices come from? Each maker's own pricing, plan or help page, in US dollars, on the check date. Most of these agents come with a general plan (a Claude plan, a ChatGPT plan, a Google AI plan, The X Subscription, a Microsoft 365 plan) rather than a separate price, so each price names the plan and its billing term, and says what the agent gets on it. Some agents also draw on credits or a shared weekly usage pool, and those limits are stated on each option. Many makers show a lower monthly price for annual billing, and some pages show only that price. Prices and plan contents change often, so check the maker's page before you buy. ### What does "Not publicly stated" mean? The maker's pages did not say. It does not mean the answer is no. ### Why are 'works in the background' and 'works inside five or more apps' not needs on this page? The pages of all ten agents say they can work in the background or on a schedule, and all ten name at least five apps or an app directory. A rule that every option meets does not tell them apart, so the page uses the narrower 'keeps working after you close the app or laptop' rule instead, and shows the apps in each option's own row. ### Is Duet on this page? Google's post of February 21, 2024 says 'Duet AI for Google Workspace is Gemini for Google Workspace', and the Google product and help pages read for this page do not use the name for a current product. A different product called Claude Duet, from Anthropic, is on a waitlist and is named in the answer above as set aside. ## Sources - https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/ (checked Nov 5, 2026) - https://ai.meta.com/muse/ (checked Nov 6, 2026) - https://blog.google/innovation-and-ai/products/gemini-app/next-evolution-gemini-app/ (checked Nov 5, 2026) - https://chatgpt.com/features/dots/ (checked Nov 6, 2026) - https://chatgpt.com/pricing/ (checked Nov 6, 2026) - https://claude.com/pricing (checked Nov 5, 2026) - https://claude.com/product/cowork (checked Nov 6, 2026) - https://docs.mistral.ai/vibe/work/connectors (checked Nov 5, 2026) - https://docs.mistral.ai/vibe/work/safety-and-approvals (checked Nov 6, 2026) - https://docs.mistral.ai/vibe/work/scheduled-tasks (checked Nov 6, 2026) - https://docs.x.ai/grok-bot/approvals-security-and-privacy (checked Nov 6, 2026) - https://docs.x.ai/grok-bot/computer-and-apps (checked Nov 6, 2026) - https://docs.x.ai/grok-bot/faq (checked Nov 6, 2026) - https://gemini.google/subscriptions/ (checked Nov 6, 2026) - https://help.mistral.ai/en/articles/347458-how-do-i-contact-support (checked Nov 5, 2026) - https://help.mistral.ai/en/articles/455207-can-i-opt-out-of-my-input-or-output-data-being-used-for-training (checked Nov 5, 2026) - https://help.openai.com/en/articles/20001530-getting-started-with-your-dot (checked Nov 6, 2026) - https://learn.chatgpt.com/docs/dots (checked Nov 6, 2026) - https://learn.chatgpt.com/docs/get-started-with-work (checked Nov 5, 2026) - https://learn.chatgpt.com/docs/pricing (checked Nov 5, 2026) - https://learn.microsoft.com/en-us/copilot/cowork/ (checked Nov 6, 2026) - https://learn.microsoft.com/en-us/copilot/cowork/cowork-faq (checked Nov 6, 2026) - https://mistral.ai/news/vibe-agent (checked Nov 5, 2026) - https://mistral.ai/pricing (checked Nov 6, 2026) - https://mistral.ai/products/vibe (checked Nov 6, 2026) - https://muse.ai/business (checked Nov 6, 2026) - https://openai.com/index/chatgpt-for-your-most-ambitious-work/ (checked Nov 5, 2026) - https://openai.com/index/introducing-dots/ (checked Nov 5, 2026) - https://support.claude.com/en/articles/13345190-get-started-with-claude-cowork (checked Nov 6, 2026) - https://support.claude.com/en/articles/9015913-how-to-get-support (checked Nov 6, 2026) - https://support.google.com/gemini/answer/13594961 (checked Nov 5, 2026) - https://support.google.com/gemini/answer/17094507 (checked Nov 6, 2026) - https://www.meta.com/help/subscriptions/1021145227643680/ (checked Nov 6, 2026) - https://www.meta.com/help/superintelligence/1047255454427887/ (checked Nov 5, 2026) - https://www.meta.com/help/superintelligence/1303670544995562/ (checked Nov 5, 2026) - https://www.microsoft.com/en-us/copilot/pricing/business (checked Nov 5, 2026) - https://www.microsoft.com/en-us/copilot/pricing/individuals (checked Nov 6, 2026) - https://www.microsoft.com/en-us/microsoft-365/blog/2026/06/16/copilot-cowork-is-now-generally-available/ (checked Nov 6, 2026) - https://www.perplexity.ai/help-center/en/articles/11187416-which-perplexity-subscription-plan-is-right-for-you (checked Nov 5, 2026) - https://www.perplexity.ai/help-center/en/articles/11521526-perplexity-tasks (checked Nov 6, 2026) - https://www.perplexity.ai/help-center/en/articles/11680686-perplexity-max (checked Nov 5, 2026) - https://www.perplexity.ai/help-center/en/articles/13837784-what-is-computer (checked Nov 6, 2026) - https://www.perplexity.ai/help-center/en/articles/13838041-how-credits-work-on-perplexity (checked Nov 6, 2026) - https://www.perplexity.ai/help-center/en/articles/13901210-computer-for-enterprise (checked Nov 5, 2026) - https://www.perplexity.ai/help-center/en/articles/2026091802-troubleshooting-connectors (checked Nov 6, 2026) - https://www.perplexity.ai/help-center/en/articles/2026102601-perplexity-teammates (checked Nov 6, 2026) - https://www.perplexity.ai/hub/pricing (checked Nov 5, 2026) - https://www.perplexity.ai/hub/products/teammate (checked Nov 6, 2026) - https://www.perplexity.ai/personal-computer (checked Nov 5, 2026) - https://x.ai/bot (checked Nov 6, 2026) Cite this page: All Things Compared, "AI agents and coworkers from the major AI makers, compared", facts checked Nov 6, 2026. https://allthingscompared.com/ai-agents-and-coworkers/