Multiplayer AI is when a company's people and its AI agents share the same work, and the agents do that work themselves, on computers of their own. Our guide to multiplayer AI explains the idea in full. This post compares the products that sell it.
Ten platforms made the list: Dock, Google Workspace coworker agents, Dust, Viktor, Claude Tag, ChatGPT workspace agents, Superconductor, QM, Microsoft Teams and Slack. For teams that want agents running ongoing business work, our pick is Dock. We build Dock, so weigh our view accordingly, and check our sources: every claim about another product links to that vendor's own page, read on November 13, 2026.
The short version: Superconductor fits software teams, QM fits teams that want to self-host, Dust fits teams that build their own agents, and Google, Microsoft and Slack fit companies already on their suites. Viktor and Claude Tag fit teams that want one shared agent in Slack.
How we compared them
Most AI at work belongs to one person. Multiplayer AI belongs to the company, and sharing is where the output is. In our November 2026 report, The State of AI Agents for Business, with data through Nov 5, the quarter of agents shared with people on the team did half of all the work. Those shared agents were also handed 58% of the work other agents started.
So we put one test to every platform, in six questions:
- Where does the agent run? A computer of its own that keeps its files and sign-ins, a short-lived sandbox, or inside a chat app.
- What happens with every laptop closed? Whether it runs on a schedule or on events with nobody online.
- Does it belong to the company? Whether the whole team can give it work, or it sits in one person's account.
- Can the team see and correct its work? Whether the work lands somewhere people can open, edit and check, with a record of who did what.
- Can agents hand work to each other? Whether one agent can pass a task to another and get the result back.
- What does a month cost? The published price, and what else you have to add up.
We read only each vendor's own pages: pricing, docs, help centers and official blogs. Where a vendor doesn't publish something, we say "not published". Prices change often, so check the linked page before you buy.
One comparison page already ranks for this term. It is published by Volition, Inc., the maker of Superconductor, and says of itself: "We are not an independent or neutral reviewer." It covers Claude Tag, Dust, QM, Superconductor and Viktor. We added Dock and the suites most teams already pay for.
Multiplayer AI platforms compared
| Platform | Where the agent runs | With every laptop closed | Shared with the team | Work the team can see | Agent to agent | Price (as of Nov 13, 2026) |
|---|---|---|---|---|---|---|
| Dock | Each cloud agent on their own always-on computer | Cloud agents run on schedules | Org or specific people; anyone shared can give it work | Shared docs and tables, edits logged by who made them | Yes; the result reports back | $40 a person a month, 5 cloud agents and an AI usage allowance |
| Google Workspace coworker agents | Google's cloud, with their own identity and persistent storage | Schedules and events (Gemini agent) | Added to a Chat space or @mentioned | Suggested edits in Docs under its own name | Yes, sub-agents | Not published |
| Dust | Dust's cloud; Computer is a temporary workspace | Schedules and triggers | Pods with shared tasks | Shared Pod files and tasks | Yes, including full handoff | Pro €24, Max €120 per seat a month, billed yearly |
| Viktor | One agent with "his own computer in the cloud" | Scheduled tasks | One Viktor shared by the workspace | Results in Slack or Teams threads | Not published (one agent) | Free up to $100 in credits; Team $100 a month for 40,000 credits |
| Claude Tag | A sandbox per conversation, discarded when idle | Routines, at most once an hour | Anyone in the Slack channel | Results in the Slack thread | Not published | No seat charge; usage balance on a Team or Enterprise plan |
| ChatGPT workspace agents | OpenAI's cloud, powered by Codex | Schedules and API triggers | Link, directory, groups, Slack | Chat, Slack and agent files | Not published | Business from $20 a seat a month, plus credits |
| Superconductor | Isolated cloud computer, paused when idle | Event-driven; time schedules not published | Every session visible to the team | Shared sessions, diffs and previews | Partly | Free; Pro $100 a month; compute credits |
| QM | A durable sandbox per person and room, in your own cloud | Crons, watches and webhooks | Channel and project agents | Chat, QM Files and web apps | Not published | Open source (MIT); you pay for your cloud and models |
| Microsoft Teams | Inside Microsoft 365; Cloud PCs for Copilot Studio agents | Status reports; autonomous agents on Cloud PCs | Per channel, Copilot-licensed members only | Status reports for team review | Not published for Channel Agent | Copilot Business from $18 a user a month; Cloud PCs $0.40 an hour |
| Slack | Inside Slack | Daily recaps and skills | Slackbot is personal; skills and Agentforce agents are shared | Shared canvases, after Confirm | Works with agents connected to Slack | Slackbot on Business+ from $15 a user a month, billed annually |
Sources for every row are linked in each section below.
The best multiplayer AI platforms
1. Dock: best for teams that want agents running the business work
Dock is where businesses hire AI agents for business work: lead research, follow-ups, CRM upkeep, reports, content and operations. People and agents work in the same channels, docs and tables.
Where agents run: each cloud agent gets their own always-on cloud computer with a persistent disk, so their files and browser sign-ins are still there the next day. A brand-new agent works from shared infrastructure until their own machine is ready. Cloud agents run on schedules and keep working when nobody is online.
Sharing: share an agent with your whole org or with specific people, and anyone it's shared with can give it work. Tag an agent in a channel and it reads the thread, does the work and replies there. Every edit to a doc or table is logged with who made it, person or agent, and agents hand tasks to each other, with the result coming back to the agent that asked.
Price: as of November 13, 2026, Starter is $40 a person a month and includes 5 always-on cloud agents and a monthly AI usage allowance. That's $8 an agent.
Strength: one plan covers the agents, their computers and an AI usage allowance. More on what a Dock agent can do.
Limit: Dock is a young company. There is no one-click Dock bot for Slack: you connect your own Slack app so messages there can wake your agents.
Best for: founders and teams of about 5 to 60 people, such as agencies and seed or Series A companies, who want agents doing sales, marketing and operations work.
2. Google Workspace coworker agents: best for companies that run on Google Workspace
Google announced coworker agents at Gemini at Work 2026: "the agent receives its own Workspace account, including an email address, calendar, Drive, and presence in your company directory." Facts as of November 13, 2026.
Where agents run: Google's cloud, with "their own persistent storage".
Sharing: colleagues work with it "by adding it to a Chat space or @mentioning it." It can "suggest an edit in the Doc and reply in the comment thread, appearing under its own name in version history," and coworker agents can "delegate and coordinate tasks."
Price: not published for coworker agents, and edition eligibility isn't in the announcement.
Strength: a real identity inside the suite your people already use.
Limit: you can't budget for it yet, because there is no published price or plan.
Best for: Google Workspace organizations happy to wait for pricing.
3. Dust: best for teams that want to build agents on company data
Dust is a platform to build, share and run agents on company knowledge and tools (dust.tt, as of November 13, 2026).
Where agents run: Dust's hosted platform. For code and files, Computer gives an agent "a safe, temporary workspace." Agents can run on a schedule or when something happens in another tool.
Sharing: in Pods, tasks are work items "that anyone (human or agent) can create, assign, start an agent on, and mark done," and files agents create are available to every member. The run_agent tool lets one agent call another, up to "a complete handoff."
Price: on the Dust pricing page, as of November 13, 2026, Free is 500 credits for life, Pro is €24 per seat a month billed yearly (€30 monthly) with 8,000 credits, and Max is €120 billed yearly (€150 monthly). Credits are charged per message and don't roll over.
Strength: model choice, with "20+ frontier and open-source models."
Limit: someone on your team builds the agents, and the workspace for code and files is temporary.
Best for: teams with a person ready to build and govern agents for everyone.
4. Viktor: best for one shared generalist in Slack or Teams
Viktor is "an AI employee that lives in Slack and Microsoft Teams" (viktor.com, as of November 13, 2026).
Where it runs: "He has his own computer in the cloud where he writes and runs code to complete tasks," and keeps "scheduled tasks that run without being asked again" (llms.txt).
Sharing: "An AI employee the company hires once and the whole team shares." It works in your connected tools and hands finished output back in Slack or Teams.
Price: on the Viktor pricing page, as of November 13, 2026, it's free to start with up to $100 in credits, and Team is $100 a month for 40,000 credits shared by the workspace.
Strength: reach. Viktor says it "connects to 9,800+ tools."
Limit: "One Viktor serves the whole workspace," so one agent holds every job.
Best for: a team in Slack or Teams that wants one generalist, billed by usage.
5. Claude Tag: best for Claude teams working in Slack
Claude Tag puts @Claude in Slack channels and threads, with Microsoft Teams on a waitlist. Facts as of November 13, 2026.
Where it runs: "in an ephemeral sandbox, not on your computer," which "is discarded when the conversation goes idle" (overview). Routines run on a schedule, at most once an hour.
Sharing: "Anyone in the channel can hand it work, and channel work bills to the organization." Results land in the Slack thread.
Price: per the overview, as of November 13, 2026, adding Claude to Slack "doesn't add a per-seat charge." Channel work draws on a usage balance an Owner funds, on a Team or Enterprise plan.
Strength: governance by channel. "Admins configure access once; everyone else just types @Claude."
Limit: the sandbox is thrown away when the conversation goes idle.
Best for: organizations on Claude Team or Enterprise that want a shared Claude in Slack.
6. ChatGPT workspace agents: best for companies already on ChatGPT Business
Workspace agents are "Codex-powered agents that automate complex workflows, run in the cloud," on ChatGPT Business, Enterprise and Edu. Facts as of November 13, 2026.
Where agents run: "Codex in the cloud, giving them access to a workspace for files, code, tools, and memory." You can run one on a schedule or trigger it through an API.
Sharing: private, by link, in the organization directory, with groups, or in Slack channels.
Price: as of November 13, 2026, a Business Standard seat is $20 a user a month billed annually ($25 monthly), and Premium is $100 ($125 monthly). Agent runs use token-based credits.
Strength: governance, with write actions set to "Always ask" by default.
Limit: the work lives in chat, Slack and agent files, and agent-to-agent handoffs aren't published.
Best for: companies already standardized on ChatGPT Business or Enterprise.
7. Superconductor: best for software teams running coding agents together
Superconductor calls itself "the multiplayer AI workspace for teams and coding agents." Facts as of November 13, 2026.
Where agents run: "Every agent runs in its own isolated cloud environment," paused when idle. Work starts from events like customer emails and meeting transcripts; time-based schedules aren't published.
Sharing: "Each agent session is accessible by everyone on the team," with shared diffs and live previews of code.
Price: on the Superconductor pricing page, as of November 13, 2026, Free is $0 with 20,000 credits for up to 4 members, and Pro is $100 a month for up to 40 members. Credits pay for cloud computer time, and you bring your own model plan or keys.
Strength: many models side by side. "Launch many agents at once, even on the same ticket."
Limit: shared documents aren't published, and the comparison page its maker publishes rates it "Does not meet" on document work.
Best for: engineers, PMs and designers who drive coding agents together.
8. QM: best for engineering-led teams that want to self-host
QM is open-source software from Y Combinator that runs agents for a team in Slack and on the web, "in your own cloud, with your own models and keys." Facts as of November 13, 2026.
Where agents run: "Every person and room gets a durable sandbox. Installed tools, files and logins carry over between turns," and "crons, process watches and signed webhooks start work when nobody is online" (agents.txt).
Sharing: "A channel or project gets its own agent with shared context." Personal agents stay isolated.
Price: no price is published. QM is MIT licensed, and you pay for your own cloud and model keys.
Strength: ownership. It runs in your cloud account, and Pi, OpenCode, Codex and Claude Code "can drive the same core."
Limit: setup is engineer-led, and your team runs and maintains it.
Best for: engineering-led teams that want full control of the stack.
9. Microsoft Teams: best for Microsoft 365 organizations on Copilot
Channel Agent, in public preview, gives each Teams channel "a dedicated agent that draws on its conversations and meetings." Facts as of November 13, 2026.
Where agents run: Channel Agent stays inside Microsoft 365. Copilot Studio agents can get a Cloud PC through Windows 365 for Agents, "provisioned per task" or always available.
Sharing: per channel, but "channel members without a Microsoft Copilot subscription can't interact with a channel's Channel Agent."
Price: as of November 13, 2026, Microsoft 365 Copilot Business starts from $18 a user a month, paid yearly. Windows 365 for Agents is $0.40 an hour in the US, plus $5 per Cloud PC a month for always-available.
Strength: enterprise management, with Intune and Microsoft 365 security applying to agent Cloud PCs.
Limit: everyone who uses the Channel Agent needs a Copilot license, and agent computers are a separate meter.
Best for: Microsoft 365 organizations with Copilot licenses and an IT team.
10. Slack: best for personal help inside Slack
Slack calls Slackbot "your personal agent for work." You can also add Agentforce and third-party agents and @mention them in channels. Facts as of November 13, 2026.
Where it runs: "natively inside your existing security perimeter" (Slackbot), with daily recaps and skills for recurring work.
Sharing: Slackbot is per person, and "your interactions are private." Teams can share skills.
Price: on the Slack pricing page, as of November 13, 2026, ongoing Slackbot access is on Business+ at $15 a user a month billed annually ($18 monthly) and Enterprise+ at $45.
Strength: context from the conversations already in Slack.
Limit: Slackbot works for one person at a time, and it runs inside Slack.
Best for: Slack customers who want each person to get help with their own work.
Where Dock is not the right pick
- Your shared work is code. Superconductor is built for teams driving coding agents together.
- You want to own and run the whole stack. QM runs in your own cloud with your own models and keys.
- Your company runs on Google Workspace or Microsoft 365 with an admin team. Their own agents sit next to your mail, files and directory.
- You want to build your own agents on company data. Dust gives you model choice and a builder.
- You want one agent living entirely in Slack. Viktor or Claude Tag will be simpler. Dock has no one-click Slack bot.
- You need compliance paperwork today. SOC 2 Type II: in progress, report to follow. We run our compliance program on Vanta, advised by Workstreet.
FAQ
What is a multiplayer AI platform? A product where a company's people and its AI agents share the same work, and the agents do that work themselves, on computers of their own. Our multiplayer AI guide covers the idea.
What is the best multiplayer AI platform? For teams that want agents running ongoing business work, we think Dock. We build Dock, so weigh our view accordingly. For code, look at Superconductor. To self-host, look at QM.
Do multiplayer AI agents need their own computer? For work that runs over days, yes. A computer that stays on keeps the agent's files and sign-ins and lets schedules fire with every laptop closed. Claude Tag discards its sandbox when the conversation goes idle, while Viktor, QM and Dock cloud agents keep a computer.
How much does a multiplayer AI platform cost? As of November 13, 2026, published prices split three ways: usage pools with no seats (Claude Tag, Viktor, Superconductor), seats plus credits (Dust, ChatGPT, Slack, Microsoft Copilot), and per person with agents included (Dock, $40 a month for 5 cloud agents). Google hasn't published a price for coworker agents.
Can AI agents hand work to each other? In some products. Dock, Dust and Google publish it clearly, and Superconductor partly. In our November 2026 report, shared agents were handed 58% of the work other agents started.
Hire agents for business work. Start for free.