Claude Dispatch

Claude Dispatch allows users to control their desktop computer via the Claude mobile app, creating a "walkie-talkie" relationship where the phone triggers tasks that are executed locally on the machine. Setup takes 30-60 seconds by scanning a QR code.

Version: 1.0.0 | Updated: 03/18/2026 | Status: Research Preview

What is Claude Dispatch?

Claude Dispatch is Anthropic's remote desktop automation feature built into the Claude mobile app. It creates a persistent, phone-to-desktop control loop where you issue a task from your phone and Claude executes it locally on your machine.

Think of it as a walkie-talkie for your computer. You're the coordinator. Claude is the operator on the ground.

All interactions run inside a single conversation thread for context continuity. The desktop side runs inside a sandboxed environment -- Claude asks permission before accessing any folder or taking any action. You approve from your phone.

How to Install

Requirements

• Claude Desktop App installed on your Mac or Windows machine
• Claude Mobile App (iOS or Android)
• Claude Max plan ($100/mo) -- currently rolling out to Pro ($20/mo)
• Your computer must stay awake with the Claude desktop app open

Setup Steps

1. Open the Claude Desktop App on your Mac or Windows machine.
2. Navigate to Settings > Dispatch.
3. A QR code will appear on screen.
4. Open the Claude mobile app on your phone.
5. Tap the Dispatch icon (or navigate to Settings > Connect Desktop).
6. Scan the QR code with your phone.
7. The connection is established in 30-60 seconds.
8. Toggle Keep Awake on in the desktop app to prevent your computer from sleeping during long tasks.

Optional: MCP Connectors

Dispatch natively supports 38 built-in connectors (Google Drive, Notion, GitHub, etc.) and 8,000+ via Zapier. You can also build custom connectors using Claude Code Skills.

To add connectors:

1. Open Claude Desktop > Settings > Connectors
2. Select a service and authenticate
3. Dispatch will now have access to that service during task execution

Use Cases

These are broken into categories by audience and complexity.

Knowledge Workers

Developers & Operators

Personal Productivity

Multi-Step / Agentic

Claude Dispatch can orchestrate multi-agent sub-tasks. A single request can fan out across browser automation, file operations, and app interactions simultaneously.

Example:

"Pull the latest Salesforce opportunity report, cross-reference it with my email inbox for any open threads on those accounts, flag ones that haven't been touched in 30 days, and draft a follow-up email for each."

This triggers: Salesforce connector + Gmail connector + file write + draft generation -- all from one phone message.

How It Works

[You on Phone] --> [Claude Mobile App]
                        |
                        v
              [Dispatch Connection (QR)]
                        |
                        v
              [Claude Desktop App - Local Machine]
                        |
                   [Sandboxed Env]
                   /      |      \
            [Files]  [Browser]  [MCP Connectors]
                        |
                        v
              [Result sent back to phone]
Click to copy

1. You send a natural language task from your phone.
2. Claude interprets the task and breaks it into steps.
3. It requests permission for any folder/app access (you approve from your phone).
4. It executes locally -- no cloud processing of your files.
5. Results surface back in the conversation thread on your phone.

Trends & Themes

Mobile-to-Desktop Remote Control

The core shift: assign complex, multi-step tasks from your phone while you're away from your desk. The persistent thread means Claude carries context across the full session -- no re-explaining.

Security via Sandboxing

Claude runs in a restricted sandbox with no default system access. Every folder and action requires explicit user approval. This is a deliberate design choice and a direct counter-positioning to tools like OpenClaw, where users configure their own access controls.

Knowledge Workers over Developers

The primary audience is non-technical users -- knowledge workers who want to delegate coordination tasks to AI without writing code. The interface is natural language. The capability is agentic.

Flat Subscription Pricing

Dispatch is included in the Max plan ($100/mo) with no per-token billing. This is a significant departure from API pricing models. Anthropic appears to be subsidizing usage to accelerate adoption -- a bet on long-term retention over short-term margin.

Research Preview Limitations

• ~50% task success rate as of v1.0
• Computer must remain awake with the desktop app running
• Single conversation thread only (no parallel sessions)
• No proactive notifications or background execution
• Limited to the Claude mobile app -- no WhatsApp, Telegram, or third-party clients

Outlier Perspectives

• One creator frames the 50% success rate as "not good enough to bet your business on" and compares it to the first iPhone -- raw potential with rough edges
• OpenClaw still wins for power users who need 24/7 server operation, 15+ messaging apps, and model-agnostic flexibility
• One suggestion gaining traction: buy a dedicated Mac Mini as an always-on Dispatch host -- leave it running, never worry about sleep states
• A broader prediction circulating: a $285B SaaS market correction as AI agents gradually replace traditional point solutions

Gaps & Blind Spots

The current version leaves several questions unanswered:

• No encryption documentation -- unclear how the phone-to-desktop connection is secured in transit
• No data retention policy -- what stays on device vs. what touches Anthropic servers
• Token consumption transparency -- flat pricing hides actual consumption, raising sustainability questions
• No platform comparisons -- no official benchmarks vs. Apple Intelligence, Windows Copilot, Zapier, or Make
• Hardware costs unaddressed -- 24/7 machine-on operation has real energy and wear costs
• No multi-thread support -- one conversation at a time limits parallel task execution
• No roadmap published -- Anthropic hasn't committed to a timeline for proactive notifications or improved success rates

Claude Dispatch vs. OpenClaw

Bottom line: Dispatch wins on accessibility and speed. OpenClaw wins on control and flexibility. The right choice depends on whether you want to configure a system or just use one.

Key Takeaways

1. Remote desktop control via mobile -- assign complex tasks from your phone while away from your desk
2. Sandboxed security model -- explicit permissions per folder and action, no silent access
3. Computer must stay awake -- the desktop app must be active; use the "Keep Awake" toggle
4. MCP connectors + Skills for extensibility -- 38 native connectors, 8,000+ via Zapier, plus custom Skills via Claude Code
5. Max plan first, Pro rolling out -- currently in research preview with ~50% reliability
6. Orchestration layer for knowledge work -- a single request can spawn multi-agent sub-tasks across browser, files, and apps
7. Flat pricing, no per-token cost -- Anthropic is subsidizing adoption; use it while the economics are favorable
8. Best starting point -- dedicate a machine (Mac Mini or similar) as your always-on Dispatch host to eliminate reliability issues caused by sleep states

Sources

1. New Claude Dispatch DESTROYS OpenClaw -- AI News Today (2.4K subs) -- 6,919 views
2. Claude Dispatch Just Dropped, And It Kills OpenClaw -- Nick Saraev (329K subs) -- 43,822 views
3. Anthropic Made Their OpenClaw -- Prompt Engineering (237K subs) -- 18,753 views
4. Claude Cowork Now Works from ANYWHERE (Dispatch) -- Brock Mesarich (76.6K subs) -- 14,548 views
5. スマホでPC動くってマジ？Claudeの新機能『Dispatch』 -- KEITO (189K subs) -- 12,033 views
6. Stop Using Openclaw. Claude Dispatch is Better -- Build Great Products (45.1K subs) -- 9,119 views