What Happened
Anthropic has quietly launched Dispatch in Claude Code Desktop—a system that allows users to assign tasks to desktop Claude from any device (including phones) and have them executed autonomously. Claude gains access to local files, configured connectors, and the browser, and can complete work while the user is away.
This is not “just another AI feature”—it’s Anthropic’s redefinition of the AI agent usage paradigm: from “you sit at the computer and talk to AI” to “you tell AI what to do, then go do your thing.”
How Dispatch Works
Task Assignment Flow
Phone/Any Device → Describe Task → Dispatch → Claude on Desktop → Execute → Result Notification- Task Description: Describe the task to be completed via the Claude App on mobile
- Route to Desktop: Dispatch sends the task to your desktop Claude instance
- Local Execution: Desktop Claude accesses local file system, configured API connectors, browser, and other tools
- Result Return: Upon completion, a notification pushes the result summary
Key Capabilities
- Local File Access: Can directly read and edit files on your computer without uploading to the cloud
- Connector Integration: Pre-configured GitHub, Slack, database, and other connectors can be called directly
- Browser Operations: Can open web pages, scrape information, submit forms
- Asynchronous Execution: No need to stay online; Claude completes tasks autonomously on the desktop
Use Case Comparison
| Scenario | Traditional Way | Dispatch Way |
|---|---|---|
| Urgent Report | Open computer → write report → send email | Describe need on phone → auto-generated on desktop → notification to check |
| Code Review | Wait to get back to desk to review PR | Describe review requirements on the go → desktop completes → result pushed |
| Data Query | Go back to office to query database | Describe query need anytime → desktop executes → result delivered |
| Document Organization | Batch process in focused time | Fragmented task assignment → desktop queues and executes |
What Dispatch truly changes is not “what AI can do,” but “when you can let AI help you work.”
Differentiation from Competitors
- vs OpenAI Codex Workflow Engine: Codex leans toward automated orchestration of code workflows; Dispatch leans toward remote triggering of general tasks
- vs Cursor MCP Apps: Cursor focuses on the development experience within the IDE; Dispatch emphasizes cross-device task continuity
- vs Traditional Automation Tools: Not pre-set if-then rules, but intelligent task understanding from natural language descriptions
Dispatch’s core advantage lies in Claude’s inherent understanding capability and the tool ecosystem already integrated into Claude Code Desktop.
Risks and Limitations
Dispatch is currently in research preview, limited to Pro/Max subscribers. Risks to be aware of:
- Security: Claude gains access to local files and connectors—users need to trust Anthropic’s security model
- Task Misunderstanding: Naturally described tasks may be misinterpreted; critical tasks still require human confirmation
- Resource Consumption: Long-running desktop Claude instances consume computing resources
- Privacy Boundaries: Users need to self-manage the privacy boundaries of local files being accessed by AI
Judgment and Recommendations
Dispatch is a key step in AI agents transitioning from “tool” to “assistant.” It solves a very practical problem: you don’t need to sit at your computer the whole time to let AI work for you.
Good for trying out: Information collection, document organization, code review, data queries—tasks where some error tolerance is acceptable.
Not recommended for: Operations involving sensitive data, irreversible modifications (deleting files, publishing content), decisions requiring precise judgment.
If you’re already a Claude Pro/Max user, try it immediately. If you’re choosing a subscription, Dispatch is a substantive differentiator between Claude and ChatGPT.