Anthropic Made Their OpenClaw

Key Concepts

• Dispatch: A new feature in Anthropic’s Claude desktop app that enables remote access to a persistent desktop AI session via a mobile device.
• Claude Co-work: The desktop-based AI environment that allows Claude to interact with system files, plugins, and browser actions.
• Open-Claude: A conceptual framework for personal AI operating systems, popularized by industry leaders like Jensen Huang, focusing on long-running, autonomous agents.
• Knowledge Work: The primary target demographic for Anthropic’s recent feature rollouts, aiming to automate complex, file-heavy, and browser-based tasks.
• Policy-based Privacy: Security guardrails designed to control how AI agents handle data and interact with external environments.

---

1. Overview of "Dispatch"

Anthropic has introduced Dispatch, a feature that bridges the gap between the Claude desktop application and the mobile app. Previously, "Claude Co-work" required the user to be physically present at their computer to supervise the AI. Dispatch transforms this into a remote-access tool, allowing users to assign tasks, retrieve files, and control browser actions from anywhere.

2. Technical Requirements and Setup

• Version Control: Users must update both the desktop and mobile applications to the latest versions to see the "Dispatch" option.
• Subscription Tiers: Currently available exclusively to "Max" plan subscribers, with a rollout to "Pro" users expected in the near future.
• Connectivity: The desktop computer must remain active and awake. The app includes a "Keep Awake" setting to prevent the computer from entering sleep mode, which is essential for maintaining the persistent connection.

3. Functionality and Use Cases

• File Management: Users can request the AI to search local directories (e.g., "Find my bank letter in the downloads folder"). The AI requires explicit permission to access specific folders, ensuring a layer of user-controlled security.
• Browser Control: Through "Claude for Chrome," the agent can interact with web services, though users are cautioned that the agent remains susceptible to phishing or malicious web content.
• Task Execution: The system functions as a continuous thread. Users can assign coding tasks or administrative duties, and the AI will execute them using the desktop's local environment.

4. Limitations and Security Considerations

• Reactive vs. Proactive: Unlike some "Open-Claude" concepts that might act autonomously, Claude currently only responds to user messages. It does not proactively initiate tasks unless it requires permission to execute a command.
• Thread Management: All interactions occur within a single, continuous conversation thread. There is currently no support for managing multiple threads or subtasks simultaneously.
• Notification Gaps: The system lacks push notifications for task completion; users must manually check the app to verify if a task has been finished.
• Scheduled Tasks: The current iteration of Dispatch does not support the "Scheduled Tasks" feature available in the standard desktop version of Co-work.
• Security Risks: Because the mobile agent has full control over the desktop (reading, moving, or deleting files), users are advised to exercise extreme caution regarding the permissions granted to the agent.

5. Strategic Context: The "Open-Claude" Vision

The industry is moving toward "Open-Claude"—a vision described by Nvidia CEO Jensen Huang as an operating system for personal AI. While Nvidia is addressing security through "Nvidia Nemo-Claude" (which enforces policy-based guardrails), Anthropic is taking a vertical integration approach by building every layer of the assistant internally. This strategy is designed to dominate the "knowledge work" sector by creating a cohesive, multi-agent ecosystem.

6. Synthesis and Conclusion

Anthropic’s release of Dispatch represents a significant step toward making AI agents truly portable and useful for professional workflows. While the product is currently limited by its single-thread architecture and lack of proactive notifications, it is highly organized compared to competitors. The primary concern moving forward is the sustainability of the current pricing model; as the company subsidizes heavy token usage for these persistent agents, users should anticipate potential price adjustments in the future. Despite these limitations, the ability to bridge desktop-level file access with mobile convenience marks a major evolution in personal AI utility.