Claude Dispatch Launch: Remote AI Control and Mobile Features

Claude Dispatch is the mobile companion to Claude Cowork. Where Cowork runs on your desktop and handles the heavy AI work — reading files, executing multi-step workflows, running agents — Dispatch lets you monitor, direct, and interact with those agents from your mobile device. The launch formalises a capability that's been in preview and adds substantial new functionality, particularly around voice input and real-time workflow status.

The Claude Dispatch launch matters for enterprise deployments for a specific reason: it addresses the "always at your desk" assumption that has limited how knowledge workers actually use Cowork. Field sales reps, executives in back-to-back meetings, and professionals who work across multiple locations can now interact with their Claude agent workflows from their phone — dispatching tasks, approving actions, and reviewing results without being tethered to a desktop.

What Claude Dispatch Is

Claude Dispatch is a mobile application (iOS and Android) that pairs with a Claude Cowork installation via QR code. Once paired, Dispatch provides a persistent connection to the Cowork session running on your desktop or in the cloud. You can see what your agents are working on, send new tasks in natural language, approve or reject agent actions that require human confirmation, and receive notifications when agents complete tasks or encounter exceptions.

The architecture is straightforward: Dispatch communicates with Cowork through a secure WebSocket connection. Your Cowork instance remains the authoritative execution environment — it retains all the context, file access, and connectors. Dispatch is the remote interface, not a separate AI system. This distinction matters for security: Dispatch doesn't have direct access to your files or connectors. It sends instructions to Cowork, which executes them within its established security boundaries.

For the full technical guide to Dispatch setup and architecture, see our Claude Dispatch setup guide. For the product overview, the Claude Dispatch guide covers features and enterprise considerations in depth.

What Launched: Full Feature Breakdown

QR code pairing (60-second setup): Open Dispatch on your phone, scan the QR code from your Cowork settings, and the connection is established. No account linking, no additional credentials. The session is tied to your Cowork authentication, not to a separate Dispatch login. Pairing is revocable from the Cowork admin console.

Natural language task dispatch: Type or dictate a task in plain language. Dispatch sends it to your Cowork session where Claude interprets and executes it using the plugins, connectors, and context already loaded. "Summarise the three documents I added to my research folder and send a draft to Sarah" works exactly as you'd expect — the files are accessed from your desktop, the summary is generated, and the draft is routed through your connected email.

Voice input: Voice-to-text integration using your device's native voice recognition, with a Claude-specific prompt interpretation layer that handles common ambiguities in spoken task descriptions. Works well for the specific kind of quick delegations that happen when you're walking between meetings: "pull the Q1 board deck and update the revenue slide with the final numbers."

Real-time agent status: See what your Cowork agents are currently working on, which tasks are in the queue, and which have completed. Status cards show task progress with enough detail to assess whether something needs your attention without requiring you to open a laptop.

Human-in-the-loop approval: For Cowork workflows configured to require approval before certain actions — sending emails, modifying files, making API calls — Dispatch surfaces those approval requests as notifications. You review the proposed action and approve or reject from your phone. This is the feature that makes Dispatch genuinely useful for professionals with high-stakes workflows: you can maintain human oversight without being at your desk.

Multi-session management: Dispatch supports multiple paired Cowork sessions — useful for professionals who run separate Cowork instances for different roles or projects, or for executives whose EA has configured a Cowork environment that they want to monitor directly.

Dispatch vs Remote Control: Key Differences

Anthropic also offers a Remote Control feature within Cowork. The distinction between Dispatch and Remote Control is worth understanding for architecture decisions.

The practical choice: use Dispatch as your default mobile companion for day-to-day delegation and approval workflows. Use Remote Control when you need full access to the Cowork desktop environment from a secondary device. They serve different use cases and most users will want both configured.

How to Set Up Dispatch in Your Organisation

Dispatch setup is deliberately simple for individual users: download the app, open Cowork settings, scan the QR code. Done in under a minute. Enterprise deployment is slightly more involved because administrators need to configure which features are available, which Cowork instances can be paired with Dispatch, and what the approval workflow permissions look like.

Enterprise IT considerations for Dispatch deployment: first, Dispatch uses the same authentication as Cowork — it inherits your SSO and MFA configuration. There's no separate Dispatch identity to manage. Second, Dispatch pairings are auditable from the Cowork admin console. You can see which devices have active pairings and revoke any session immediately. Third, the human-in-the-loop approval notifications require push notification permissions on the mobile device — make sure your MDM configuration allows this for the Dispatch app if you're managing employee devices.

Enterprise Use Cases

Executive briefing and decision support: An executive's EA configures a Cowork session that continuously monitors news feeds, email, and calendar for items requiring the executive's attention. Dispatch surfaces a daily briefing summary and flags items needing approval or response. The executive reviews on their phone between meetings, dispatches responses or directs follow-up via voice, and never has to open a laptop for routine information management.

Field sales rep workflow: A sales rep configures Cowork to handle proposal generation, CRM updates, and competitive research. During field visits, they dispatch tasks via Dispatch ("update the Acme opportunity status and generate a follow-up email after my meeting") and receive results before they've finished walking back to their car. The CRM stays current without manual data entry.

Legal review approval chains: A law firm configures Cowork to perform initial contract review and flag clauses for partner approval. Partners receive Dispatch notifications when contracts are ready for sign-off, review the summarised risk flags on their phone, and approve or request escalation. The review cycle compresses from days to hours without sacrificing partner oversight.

IT incident response: An IT operations team configures Cowork agents to monitor system health and generate incident reports. On-call engineers receive Dispatch notifications for new incidents, can review the agent's preliminary analysis, and dispatch investigation or remediation tasks directly from their phone. Escalation decisions happen faster because the context is immediately available.

Security Model and Enterprise Governance

The Dispatch security model is built on the principle that Dispatch is an interface to Cowork, not a separate system with its own access to enterprise data. The key security properties: Dispatch doesn't store or process enterprise data directly — it proxies through Cowork, which has the authenticated connectors and access controls. Cowork-level permission restrictions apply to tasks dispatched via the mobile app exactly as they apply to desktop sessions.

Device security is the primary consideration. A compromised mobile device with an active Dispatch pairing is a compromised Cowork interface. Organisations should ensure Dispatch is included in MDM policies, that device encryption is enforced, and that Dispatch pairings are automatically revoked if a device is reported lost or stolen. The admin console in Cowork provides immediate pairing revocation — this should be in your device management runbooks.

The human-in-the-loop approval feature actually improves the security posture for high-stakes workflows by requiring explicit human confirmation for consequential actions. This transforms what might have been a fully autonomous agent action into one that requires a human decision — reducing the blast radius of any prompt injection or mis-specification issue in the underlying workflow.

What's Coming Next

The Dispatch launch is explicitly framed as the foundation for a mobile-first Claude experience, not a complete product. Several features are on the near-term roadmap: widget support for iOS and Android home screens (one-tap task dispatch without opening the app), deeper integration with calendar apps for context-aware task suggestions, and expansion of the multi-session management features for organisations deploying shared Cowork environments.

Longer term, Dispatch is the obvious interface layer for a world where AI agents are running continuous background workflows. As Cowork evolves to support more persistent, longer-running agent tasks — processes that run over hours or days, not minutes — Dispatch becomes the monitoring and control plane for that ongoing work. The mobile interface is the right place for that kind of episodic human oversight.

For organisations planning their Claude AI strategy over the next 12-24 months, Dispatch should be in your Cowork deployment architecture from day one. Configuring it correctly at launch is substantially easier than retrofitting it later when your agents have become operationally embedded. Our Cowork deployment service includes Dispatch configuration as a standard component.