Claude Dispatch is the connective layer that turns Claude from a single chat window into a coordinator for work happening across your devices. Instead of asking one assistant to do one thing in one place, Dispatch acts as an orchestrator: it receives a request, decides what kind of work it involves, and routes that work to the right execution environment on your own machine. The result is a system where you can start something from anywhere and let it run where it makes the most sense.

At its core, Dispatch is a router. When you send it a request, it interprets your intent and hands the job off to one of two kinds of sessions. The first is a task session, a lightweight, self-contained run designed to carry a discrete piece of work from start to finish — drafting a document, summarizing a thread, pulling together research, or completing any well-scoped assignment. The second is a Code session, a full coding environment running on your computer with access to your files, your repositories, and your tools. Code sessions are where heavier, hands-on engineering work lives: editing a codebase, running tests, debugging, or shipping a change. Dispatch’s job is to look at what you asked for and send it down the right path, so you don’t have to choose the environment yourself every time.

What makes this genuinely useful is where it lives. Dispatch is available from the desktop, where most deep work already happens, and from the phone app through a dedicated Dispatch tab. That means the orchestrator is always within reach. You can be away from your desk, open the app, and kick off a task or a coding session that executes back on your machine. The work runs in the right place — with the right context, files, and compute — even though you triggered it from your pocket. When you return to your desktop, the sessions are there, with their progress and output waiting.

This separation of where you ask from where it runs is the heart of the idea. A request is just an intention; Dispatch is responsible for matching that intention to capability. Because task sessions and Code sessions run on your own hardware, they keep access to the real environment your work depends on, rather than a sandbox that only approximates it. And because the orchestrator sits above them, you get a single, consistent way to start, monitor, and pick up work no matter which kind of session ends up doing it.

The practical payoffs show up quickly. You can dispatch a long-running coding task before you leave the house and check on it from your phone during the commute. You can fire off several independent tasks at once and let them progress in parallel rather than waiting on a single conversation. You can keep a clean separation between quick, throwaway jobs and serious engineering work without juggling different apps or losing track of what is running where. And because everything flows through one entry point, there is far less friction between having an idea and having something act on it.

For anyone who works across more than one device — which is to say, almost everyone — Dispatch closes the gap between where inspiration strikes and where execution happens. It treats your machine as the engine and your phone or desktop as the remote, with an intelligent dispatcher in between deciding how best to get the work done. This page is a simple test of that flow: a small, deliberate example created to confirm that a request can travel from a prompt to a finished result, routed and executed exactly as Dispatch is designed to do.