Claude Cowork vs Dust: Shared Task Ownership
Compare Claude Cowork and Dust on task assignments, event triggers and reusable procedures before choosing where team work should live.
Founder, Task Machine
Should Claude Cowork or Dust hold the work when a developer hands a job to another team member? Dust's Pod task guide documents assignees and linked agent conversations. Cowork retains session progress and offers an editable task list, but its docs do not establish the same shared ownership record.
That distinction matters after the initial request. For a research brief you personally commission and review, a session may be enough. For a request that moves between a person and an agent, evaluate the assignment and intervention path before comparing the quality of the final document.
Cowork keeps delegation close to the conversation
Claude Cowork handles multi-step knowledge work across files and connected tools. Its getting-started instructions describe analyzing the request, making a plan and breaking complex work into subtasks. You review the approach before allowing execution, and manual mode pauses for permission requests.
The Productivity task-management skill maintains a task list that people and Claude can edit. That is useful for keeping a list beside delegated work. It does not establish work items that either people or agents own with their own discussion and history, so test the handoff you need without assuming the list supplies it.
Dust makes the assignment visible in a Pod
Dust's Pod task documentation describes assignees and Open, In progress and Done states. People can complete tasks, and agents can pick them up and mark them done. Starting agent execution opens a linked conversation where you can follow the work and intervene.
That is a concrete match for a small team sharing requests and company context. It still leaves a boundary around coordination: Dust's initiative example describes progress summaries, and its Run agent guide describes delegation, but its public docs do not describe an assigned project, goal or team lead responsible for reviewing and coordinating the work. An assignee and a persistent lead are different responsibilities.
What starts the next occurrence?
Cowork's schedule guide supports hourly, daily, weekly, weekday and on-demand work, with past and upcoming runs to review. Its public Cowork pages do not describe outside events or webhooks starting tasks. A weekly report fits that published schedule model without needing an event connection.
Dust's trigger guide documents schedules and webhooks, with each run opening a conversation that can belong to a shared Pod. If the report must start when an outside service sends an event, Dust publishes a direct path. Decide whether the job is genuinely event-driven before adding trigger setup to an otherwise adequate schedule.
Webhook capacity also affects the handoff. Dust's rate-limit documentation says triggers default to 42 runs per sliding 24 hours, editors can change the limit, and requests arriving at the limit are dropped rather than queued. Check the intended burst pattern before routing incoming requests through that trigger.
Reusable procedures leave step selection with the agent
Cowork's Skills documentation describes saved procedures and scripts. Its published ticket-triage skill contains ordered categorization and routing instructions, but the docs do not establish an enforced multi-step process that repeats the same execution path. A written sequence is useful without proving that every run must follow it.
Dust's skills guide also packages reusable instructions, and skills can reference other skills. The agent decides which skills to activate. Dust's public docs do not describe a native fixed-step workflow with defined branches that repeats the same path, so neither product's published reusable instructions should be treated as that execution contract.
For code, both publish more specific checks. Cowork's Engineering review skill can compare a pull request with requirements and inspect test results. Dust's GitHub tools support reviewer agents checking internal coding rules. These are code-review capabilities, not published acceptance checks on every general task delivered through either product.
Execution and payment use different units
Cowork's plan guidance documents cloud and local virtual-machine sessions for Team and Enterprise. Its individual-plan guidance says new Pro, Max and Ultra tasks run in the cloud only, so do not assume every account has the same local option. Its pricing page lists Pro at $20 monthly or $200 annually, Max at $100 monthly and Ultra at $200 monthly, with Cowork drawing from the paid Claude allowance.
Dust's Computer documentation describes a temporary workspace for files and code. Its public docs do not establish running the agent runtime on your computer or infrastructure. Dust's pricing page lists Pro at €30 per seat monthly or €24 per month billed yearly with 8,000 monthly credits, excluding VAT. A Claude plan allowance and Dust seat credits cover different units, so use a representative job to assess their fit.
Claude Cowork and Dust, point by point
The table gives both products' answers to the same questions, with a source and check date for each answer. Read a row's explanation alongside its label, especially where a published capability covers a narrower case.
Both columns answer the same questions from each company's public pages. Yes means the company publishes the capability the question asks about, Partly means something narrower, No means it says it does not offer it, and Not published means its public pages do not describe it. Every answer links to its source and shows when we last checked it. The oldest check on this page was November 13, 2026.
| What you compare | Claude Cowork | Dust |
|---|---|---|
| Goals |
Claude Cowork
Partly
You can give Claude an outcome and schedule a weekly marketing readout with metrics. Its docs do not describe a goal record with an owner tracking a measurable target. |
Dust
Partly
A Pod can hold an initiative's goal and success criteria, with scheduled progress summaries. Its docs do not describe a named goal owner tracking a measurable target and acting on its changes. |
| Leads |
Claude Cowork
Partly
Claude coordinates parallel sub-agents within a task. Its project docs describe sharing and context, not a persistent lead role reviewing and coordinating a project or team. |
Dust
Partly
Agents can coordinate project updates and delegate work to specialist agents. Its docs do not describe an assigned project, goal or team lead role responsible for reviewing and coordinating the work. |
| Tasks |
Claude Cowork
Partly
Cowork sessions retain progress and results, and the Productivity plugin maintains a task list people and Claude can edit. Its docs do not establish work items owned by either people or agents with their own discussion and history. |
Dust
Yes
Pod tasks have assignees and Open, In progress and Done states. People can complete them, agents can pick them up and mark them done, and agent execution opens a linked conversation where you can intervene. |
| Recurring work |
Claude Cowork
Partly
Scheduled tasks run hourly, daily, weekly, on weekdays or on demand, with upcoming and past runs to review. Its Cowork docs do not describe outside events or webhooks starting tasks. |
Dust
Yes
Schedule triggers run agents at times you describe, and webhook triggers respond to outside events. Each run creates a conversation, optionally in a shared Pod. |
| Workflows |
Claude Cowork
Partly
Skills save procedures and scripts, and the ticket-triage skill defines ordered categorization and routing steps. Its docs do not establish an enforced multi-step process that runs the same way each time. |
Dust
Partly
Skills package reusable instructions and can reference other skills, while agents delegate through Run agent. The agent decides which skills to activate. Its docs do not describe a native fixed-step workflow with defined branches that repeats the same execution path. |
| Autonomy |
Claude Cowork
Yes
Connector tools can be set to Always allow, Needs approval or Blocked. Manual mode pauses for permission requests you allow or deny in the task. |
Dust
Yes
Admins configure MCP tool approval levels, including always asking and never asking. Agents pause for approval in conversations, and actions requiring confirmation from an emailed request send private Accept and Reject links to the sender. |
| Learned autonomy |
Claude Cowork
Not published
Its public Cowork, permissions, memory, plugin and release documentation does not describe proposals to change an agent's authority based on approval history. |
Dust
Partly
Medium-stake tools retain per-agent approval history, and self-improving skills propose instruction and tool changes from conversation feedback. Its docs do not describe proposals to change action permissions based on approval history. |
| Memory |
Claude Cowork
Yes
Cloud Cowork tasks share memory with chat, and projects remember context for future tasks. You can inspect saved memories in Settings and edit or delete individual entries. |
Dust
Yes
Agent Memory keeps user-specific information across conversations. You can inspect and delete saved memories in the agent's Memory tab. Each agent has separate memories for each user. |
| Planning |
Claude Cowork
Yes
Claude analyzes your request, creates a plan and breaks complex work into subtasks. The getting-started instructions tell you to review Claude's approach before letting it run. |
Dust
Yes
For complex questions, @deep-dive creates a research plan and consults a planning agent to review and improve it before delegating to specialist agents. You can follow the research reasoning as it runs. |
| Intake |
Claude Cowork
Partly
The Customer Support plugin uses connected ticket history and correspondence, categorizes incoming tickets and recommends a team or queue. Its published workflow describes routing advice, not assigning an incoming request to an owner or agent. |
Dust
Yes
Webhook sources receive events such as new tickets and route them to an agent's configured trigger. Workspace members can also forward email to [email protected]. Email from non-members is not processed. |
| Connectors |
Claude Cowork
Partly
Claude offers a connector directory and custom remote Model Context Protocol servers, with plugins for connected apps. Its Cowork docs do not describe webhook or event triggers that start work. |
Dust
Yes
Dust offers built-in MCP tools, remote MCP servers and custom servers at a public URL. Built-in and custom webhook sources can start agents from events in connected services. |
| Models |
Claude Cowork
Partly
You can select the model for a scheduled task, and the unified experience retains a model picker. Its docs do not describe automatic model selection per task or step. |
Dust
Partly
You choose a specific model or Basic, Standard or Premium. Auto uses a maintained ordered list with availability fallback. Dust explicitly says it does not select a different model based on each task. |
| Cloud or your own computer |
Claude Cowork
Yes
Team and Enterprise document cloud sessions and local virtual-machine sessions. New Pro, Max, and Ultra tasks run in the cloud only, so the local option depends on your plan and experience. |
Dust
Partly
Computer gives agents a temporary workspace for code and files. A preview supports tools executing in custom client applications. Its docs do not establish an option to run the agent runtime itself on your computer or infrastructure. |
| Your own subscription |
Claude Cowork
Yes
Cowork is included in paid Claude plans and uses the same plan allowance, so an existing Claude subscriber can delegate work without a separate Cowork subscription. |
Dust
Yes
Dust's Gemini 4 Argon announcement explicitly includes workspaces that bring their own API key. That announcement does not specify eligible plans or the full set of supported providers. |
| Pricing |
Claude Cowork
Yes
Pro costs $20 monthly or $200 annually. Max costs $100 monthly and Ultra $200 monthly. Enterprise costs $20 per seat monthly, billed annually with a 20-seat minimum, plus usage at API rates. |
Dust
Yes
Business offers Free seats with 500 lifetime credits, Pro at €30 per seat monthly or €24 per month billed yearly with 8,000 monthly credits, and Max at €150 monthly or €120 per month billed yearly with 40,000 monthly credits. Prices exclude VAT. Enterprise is custom. |
| Correctness |
Claude Cowork
Yes
The Engineering plugin checks a pull request against its stated requirements and can inspect CI status and test results when source control is connected. These are optional plugin checks, rather than automatic checks on every delivered file. |
Dust
Yes
Dust documents PR reviewer agents that review pull requests against internal coding rules. This is a defined-criteria check for code reviews, rather than a published acceptance check on every Pod task. |
For software work
| What you compare | Claude Cowork | Dust |
|---|---|---|
| Code review rules |
Claude Cowork
Yes
The Engineering plugin reviews code for security, performance, correctness and maintainability. With a knowledge-base connection, it checks changes against team coding standards and style guides. |
Dust
Yes
GitHub tools let you build PR reviewer agents that review pull requests based on your internal coding rules. |
| Adversarial review |
Claude Cowork
Not published
Its public Cowork and Engineering plugin docs do not describe an independent second review of risky changes. Published sub-agent coordination and a code-review skill do not establish that review contract. |
Dust
Not published
Its public docs describe no independent second review of risky code changes. GitHub reviewer agents and sub-agent delegation do not establish that review process. |
| Risk rating |
Claude Cowork
Partly
The code-review skill reports issue severity and an Approve, Request Changes or Needs Discussion verdict. Its docs do not describe an overall risk rating for the exact change before merge. |
Dust
Not published
Its public docs describe no pre-merge risk rating for a code change. Published GitHub review tools and engineering use cases do not establish a risk score. |
| Pull requests |
Claude Cowork
Partly
The Engineering plugin accepts a pull-request link, fetches its diff through connected source control and reviews it. Its docs do not establish agents revising the branch in response to pull-request review feedback. |
Dust
Partly
GitHub tools can search, update, comment on and create pull requests, and agents can review them. Its docs do not establish agents revising a code branch in response to pull-request review feedback. |
A fact we cannot find or cannot check is left out, and prices can change after the date shown. Task Machine publishes this comparison and is an interested party. Claude Cowork and Dust are named only to identify them, and there is no relationship between the companies. Has something changed? Tell us.
Who each one suits
Pick Claude Cowork if you want to delegate file work within a paid Claude plan and review its approach in conversation. Its task list remains a narrower published work model than assigned Pod tasks, and its public pages do not describe webhook-started Cowork jobs.
Pick Dust if the deciding requirement is a shared Pod assignment with a linked agent conversation and schedule or event triggers. Its fixed-step workflow contract and customer-hosted agent runtime are not established in the public docs. Keep the webhook drop behavior in the evaluation if requests arrive in bursts.
Where Task Machine fits
Task Machine answers the ownership and execution-path questions with Tasks owned by a person or agent and Workflows with defined steps, branches and verifier checks. A Task retains its plan, discussion, decisions and run history, while failed verifier checks can return work with feedback. That fits a team whose requirement combines a shared assignment with an explicit result-checking path.
Task Machine is the wrong pick when the main requirement is simply to keep delegated file work inside a paid Claude allowance or continue using Dust's existing Pod conversations. Compare the Claude Cowork ownership details and the Dust ownership details, then map one person-to-agent handoff against both records before moving it.