Top 200 ReposEXPLORER
Browse agents, categories and pages
Back to results
Planned, not doneThe source describes an intention, not a finished result.

Hand off research and document tasks to the Dispatch agent in Claude Cowork

Anthropic’s Cowork docs describe Dispatch, a background agent in Claude Desktop that splits one request into child tasks, runs each as its own session and forwards permission prompts to you, even from your phone.

The problem

How it works

Cowork runs inside Claude Desktop on the same agent architecture as Claude Code. It reads and writes local files, can work with Claude in Chrome, splits large jobs across sub-agents and can produce spreadsheets with working formulas, slide decks and formatted documents. Dispatch sits on top of it: you describe the result you want in one conversation, and the Dispatch agent breaks it into child tasks. Coding tasks go to a Claude Code workspace you have already set up, and knowledge work goes to a Cowork project. Each child task appears in the sidebar with a status such as running, awaiting input, completed or error, and you can open its transcript and files. With Claude Desktop left open, you can also start Dispatch work from the Claude mobile app.

Results and limitations

This is product documentation, not a reported result. The docs give a sample brief but no success rates, timings or examples of finished work.

Before you try it

Dispatch needs a Pro, Max or Ultra plan and the latest Claude Desktop on macOS or Windows, and phone hand-offs only work while your computer stays open, awake and online. A permission request you do not answer within ten minutes is denied and the task carries on without that step, so read the transcript for skipped actions. Child tasks cannot start further children of their own. Cowork uses the connectors, skills and plugins enabled on your claude.ai account, not the ones in a local Claude Code setup. Results depend on your setup and were not independently reproduced by Top 200 Repos.

Tools and repositories

No ranked repository is involved. This use case still stands on its own source.

Tools
Claude Desktop with Cowork · Cowork projects · Claude mobile app (optional)
Prerequisites
A Claude Pro, Max or Ultra plan and the latest Claude Desktop app on macOS or Windows.

Try it yourself: Brief a background task with a clear finish line

Fill the fields to adapt the prompt. Empty fields stay as visible [placeholders].

Preview
Using the [Cowork project] project, research [Research question] and write [What to produce] in that project’s folder. Before you start, list the child tasks you plan to run and the sources or files each will use. Cite every source you rely on and flag anything you could not verify. Do not send messages, delete files or change anything outside the project folder; if a step needs that, stop and ask me. When you finish, tell me what you produced, what is missing and where the files are.

Answer permission prompts promptly, because unanswered requests are denied after ten minutes.

Sources

  1. Run tasks in the background with Dispatch (opens in a new tab)

    Project documentation · claude.com · Undated · Source checked 12 Nov 2026, 07:57 UTC

    What this source supports
    • Dispatch breaks a described outcome into child tasks that run as separate Cowork or Code sessions
    • Coding work runs in an existing Code workspace and knowledge work runs in a Cowork project
    • Child tasks show states such as running, awaiting input, completed and error, and cannot start further children
    • Permission prompts are forwarded to the user and denied automatically after ten minutes
    • Requires a Pro, Max or Ultra plan and the latest Claude Desktop on macOS or Windows
    • Tasks can be started from the Claude mobile app while the desktop stays open, awake and online
  2. Overview (Cowork) (opens in a new tab)

    Project documentation · claude.com · Undated · Source checked 12 Nov 2026, 07:54 UTC

    What this source supports
    • Cowork uses the same agent architecture as Claude Code inside Claude Desktop
    • Reads and writes local files without manual uploads or downloads
    • Splits complex work into parallel sub-agent workstreams
    • Creates Excel files with working formulas, PowerPoint decks and formatted documents
    • Loads the connectors, skills and plugins enabled for the claude.ai account, not the local ~/.claude directory
Report a problem

Planned, not done

Delegate errands to Meta’s Muse, with approval before it sends or buys

Meta’s Muse runs in its own secure cloud VM, takes requests in WhatsApp or the Muse app, and asks before sending email or buying; a separate Sentinel agent must approve anything it sends to the internet.

about.fb.com · 8 Sept 2026
Reported by source

Keep a resident desktop agent for research notes and blog work

Raj Chettri runs Nous Research's Hermes Agent on his desktop for scheduled research summaries, note filing, quick scripts and his blog's writing pipeline.

rajchettri-com.pages.dev · 25 May 2026NousResearch/hermes-agent
Reported by source

Answer technical questions by letting async agents run experiments

Simon Willison sends research questions to cloud coding agents such as Claude Code for web and Codex Cloud, which write and run experiments in a dedicated repo and open a PR.

simonwillison.net · 6 Nov 2025anthropics/claude-code, openai/codex
Reported by source

Build a cited Q&A knowledge base over your own documents

Self-host RAGFlow, upload PDFs, Office files and scans, and ask questions that come back with traceable source chunks.

github.cominfiniflow/ragflow
Planned, not done

Hand CRM updates and follow-up drafts to Grok Bot agents with approval steps

Grok Bot launched in beta in August 2026: always-on agents with their own cloud computer that sign in to your apps, learn routines by watching you, and come back when something needs your approval.

x.ai · 11 Aug 2026
Reported by source

Let a coding agent run overnight ML training experiments

Point a coding agent at autoresearch so it edits a small GPT training script, runs short timed experiments and keeps what improves.

github.comkarpathy/autoresearch, anthropics/claude-code, openai/codex