Claude Dispatch: Anthropic Turns Your Mac Into an Autonomous Agent You Control From Your Phone
Claude Dispatch turns your Mac into an autonomous AI agent controlled from your phone. Local-first, sandboxed, human-in-the-loop approval on every action.
Your Mac is on. You're not in front of it. Until now, that was an idle resource — a machine burning power and producing nothing. Claude Dispatch just changed that.
Announced on March 16, 2026 by Felix Rieseberg (Anthropic), Dispatch is a new feature inside Claude Cowork, Anthropic's macOS app. The idea: you send a mission from your phone. Claude executes it on your Mac. You come back — it's done.
Your Mac Works. You Don't.
Imagine texting your computer to delegate a task. That's exactly what Claude Dispatch does.
A persistent conversation with Claude runs continuously on your Mac. From your phone — iOS or Android — you send an instruction. Create a report from internal dashboards. Merge PDFs. Find a better seat on an upcoming flight. Claude handles it.
As Felix Rieseberg puts it: "One persistent conversation with Claude that runs on your computer. Message it from your phone. Come back to finished work."
No developer skills required. No terminal. This is an AI agent (a program that executes tasks autonomously) accessible to everyone.
And the experience? Felix Rieseberg sums it up: "It feels pretty magical to give Claude a mission on my computer and getting occasional updates."
How Dispatch Works in 5 Steps
Claude Dispatch follows a simple and secure flow.
1. A persistent conversation runs on your Mac. Claude Cowork must be active and your Mac powered on. That's the foundation.
2. You send a mission from your phone. A text message is enough. "Merge the 3 PDFs on my desktop and email the result."
3. Claude executes inside a sandbox. A sandbox is an isolated environment that prevents the agent from accessing unauthorized resources. Your files, browser, and tools are reachable — but contained.
4. You approve before every action. Claude does nothing without your go-ahead. You receive updates on your phone and validate each step.
5. You come back — it's done. The result is waiting on your Mac.
As Felix Rieseberg emphasizes: "Everything Claude can do on your computer — files, browser, tools — are reachable from wherever you are."
Privacy-First and Human-in-the-Loop: Anthropic Plays It Safe
Claude Dispatch is local-first: your files never leave your machine. No data goes to Anthropic's servers. That's a fundamental difference from cloud-based AI agents like OpenAI Operator or Google Gemini Agent.
The human-in-the-loop principle (mandatory human oversight) is non-negotiable. Claude asks for your approval before every action. No silent execution. No surprises.
This is a deliberate choice by Anthropic. Where agents like Devin (Cognition) or Manus Desktop push toward maximum autonomy, Anthropic refuses to cross the line. Safety before speed.
| Agent | Execution | Local Files | Human Approval | Mobile Control |
|---|---|---|---|---|
| Claude Dispatch | Local (Mac) | ✅ Yes | ✅ Required | ✅ Yes |
| OpenAI Operator | Cloud | ❌ No | ⚠️ Partial | ❌ No |
| Google Gemini Agent | Cloud | ❌ No | ⚠️ Partial | ⚠️ Limited |
| Devin (Cognition) | Cloud | ❌ No | ⚠️ Optional | ❌ No |
| Microsoft Copilot | Cloud + local | ⚠️ Partial | ⚠️ Partial | ⚠️ Limited |
Claude Cowork: The Foundation Dispatch Builds On
Dispatch doesn't exist in a vacuum. It's a feature of Claude Cowork, the macOS app Anthropic launched in January 2026.
Cowork targets non-developers. Where Claude Code is built for devs, Cowork automates everyday tasks: expense reports, file conversions, reports, exports.
Notable detail: Cowork was entirely coded in 10 days by Claude itself. An AI agent that builds the tool that will host another AI agent. Full circle.
The first hands-on from MacStories is honest. Connectors work. Results are still 50/50. It's slow. But the direction is clear: "A step in the right direction."
Dispatch is in research preview (experimental version). Available now for Max subscribers. Pro subscribers will get access within the next few days.
What This Signals: Your Machine as an AI Execution Layer
After cloud agents — OpenAI Operator, Google Gemini — Anthropic takes a different path. The bet: local execution.
Every Mac that's powered on becomes an AI worker. No data center needed. No cloud required. Your machine, your files, your control.
This is a paradigm shift. Agentic AI tools ran in the cloud or inside an IDE until now. Dispatch brings intelligence back to the individual machine.
And it's a signal for the entire industry. If a consumer-facing AI agent can turn a personal Mac into an autonomous employee controlled remotely, the very concept of "productivity" changes. We're no longer talking about tools. We're talking about personal infrastructure.
In summary:
- Claude Dispatch, announced March 16, 2026, turns your Mac into an autonomous agent controlled from your phone.
- A persistent conversation with Claude runs locally — your files never leave your machine.
- You approve every action before it executes — Anthropic maintains human-in-the-loop control.
- Use cases: reports, file management, browser, tools — everything Claude can do on your Mac is reachable remotely.
- Available now for Max subscribers, with Pro access coming within days.
Dispatch is the concrete, consumer-facing version of a simple idea: your Mac on your desk, your phone in your pocket, a persistent conversation running — your idle resources aren't idle anymore. This isn't productivity. It's personal infrastructure. And according to this full analysis, it's just the beginning.
