Update March 24, 2026: Anthropic launched computer use in Cowork and Claude Code. Claude can now control your mouse, keyboard, and screen to complete tasks. This article has been updated to reflect this development.
In the AI world, much of the attention in recent weeks has gone to OpenClaw: the open-source AI agent platform that collected more than 250,000 GitHub stars in record time and was called "the next ChatGPT" by Nvidia CEO Jensen Huang just last week. The lobster emoji 🦞 has become the unofficial symbol of a new generation of AI assistants that don't just answer questions but actually do things on your computer.
What's getting far less attention — while being much more accessible and therefore more interesting for most people — is that Anthropic, the company behind Claude, has launched virtually identical functionality during the same period. It did so through an almost daily series of individual updates that, taken together, look remarkably similar to what OpenClaw offers. Let's break that down.
What is OpenClaw, exactly?
For those who missed it: OpenClaw is a personal AI assistant that runs on your own hardware. What makes it special is that you communicate with it through apps you already use — WhatsApp, Telegram, Slack, Discord, Signal, iMessage — and that the platform can execute tasks autonomously. Think processing emails, managing your calendar, organizing files, and even controlling your browser.
The core features that made OpenClaw popular:
Cron jobs: tasks that run automatically at set times, like a morning briefing
Webhooks: external triggers that activate the assistant (for example, an incoming email)
Skills: modular capabilities you can add, share, or have the AI build for you
Browser control: the AI can operate websites, fill in forms, and retrieve data
Multi-channel messaging: you send a message via WhatsApp and the assistant executes it on your computer
Persistent memory: the assistant remembers context across conversations
Local execution: everything runs on your machine, not in the cloud
The result is an AI that feels like a digital colleague on standby 24/7. No wonder it went viral.
The backstory: from Clawdbot to OpenClaw — and why it matters
OpenClaw was originally called Clawdbot. The name was a play on Claude, Anthropic's AI model, combined with "claw" (lobster claw). The project was built on top of Claude, and its creator, Peter Steinberger, made no secret of that. The mascot was a lobster and the name a nod to the underlying model.
Anthropic wasn't pleased. On January 27, 2026, the company sent a legal letter demanding a name change. The name "Clawd" in Clawdbot was too close to their registered trademark "Claude."
Steinberger cooperated. That same day, the project was renamed to Moltbot — a reference to the molting of a lobster. But in the minutes that the old GitHub name and X account were released, scammers struck. The original @clawdbot account with more than 60,000 followers was hijacked and used to promote a fake cryptocurrency ($CLAWD). So Steinberger had to find the now-definitive name: OpenClaw. The lobster stayed; the reference to Claude disappeared.
But Anthropic continued to push back. It blocked the ability for tools like OpenClaw to log in to Claude subscriptions via OAuth. And on February 14, Steinberger announced he was moving to OpenAI, with OpenClaw placed under an independent open-source foundation with OpenAI as financial sponsor.
What has Anthropic launched?
Between January and March 2026, Anthropic rolled out an impressive series of features. While each individually appears to be a minor product improvement, together they form what makes OpenClaw so compelling.
Cowork
The foundation of everything. Cowork gives Claude direct access to files on your computer through folder access within a secured sandbox. Claude can read, edit, create, and organize files. It runs as a virtual machine on your desktop and executes multi-step tasks autonomously. Compare that to OpenClaw's local gateway that manages files and scripts.
Plugins and Marketplace
OpenClaw has its Skills and ClawHub — a registry where the community shares capabilities. Anthropic responded with a full plugin marketplace. Eleven official plugins for functions like sales, finance, marketing, legal, and product management. Each bundles skills, slash commands, MCP connectors, and sub-agents into a single package.
For businesses, Anthropic went a step further: private marketplaces where IT administrators can distribute plugins to specific teams, including automatic installation and access control.
Scheduled Tasks and the cloud
The OpenClaw equivalent of cron jobs. Using the /schedule command in Cowork, you can set up recurring tasks: daily email summaries, weekly reports from Google Drive, or monthly file cleanups. Claude Code received similar functionality with the /loop command, including support for cron expressions.
And then the latest step: as of today (March 21), you can also run scheduled tasks in the cloud. That means your computer doesn't even need to be turned on. This solves one of the biggest limitations of both OpenClaw and the earlier Cowork variant — with both, your machine had to be active to execute scheduled tasks.
Dispatch
This is the feature that triggered the direct comparison with OpenClaw. Dispatch makes it possible to send tasks from the Claude app on your phone to your desktop computer. You scan a QR code, pair your phone, and can then give instructions from the train or the couch that Claude executes at your desk. Dispatch is now available in both Cowork and Claude Code.
With the recent addition of computer use, Dispatch becomes even more powerful: Claude can now not only edit files but also control your screen while you're on the go. Think of a morning briefing being assembled while you're on the train, or changes in your IDE being made — including tests and a pull request.
With OpenClaw, you do exactly the same thing, but via WhatsApp or Telegram. Dispatch is the polished, closed alternative, making it accessible to far more people.
Computer use (since March 24)
The most recent addition and perhaps the most direct catch-up to OpenClaw's browser control. Claude can now operate your mouse, keyboard, and screen to complete tasks. No connector available for a particular service? Claude simply opens your browser, clicks, scrolls, and fills things in. Combined with Dispatch, the picture is complete: you send a task from your phone, Claude operates your Mac to carry it out. The feature is available as a research preview for Pro and Max subscribers, currently on macOS only. Anthropic has built in safeguards: the model explicitly asks for permission, certain apps are blocked by default, and automatic prompt injection detection is running.
Claude Code Channels
This might be the most direct OpenClaw counterpart in terms of experience. You can now control Claude Code via Telegram and Discord — exactly the messaging apps that made OpenClaw big. You send a message via Telegram, Claude Code executes it, and sends the result back.
The difference from Dispatch: Channels targets developers who use Claude Code, while Dispatch is intended for knowledge workers in Cowork. But the principle is identical to OpenClaw: reaching your AI through the chat app you already have open.
Auto-memory
One of OpenClaw's most praised features is persistent memory — the assistant remembers your preferences, project context, and work patterns in local Markdown files. Claude answered this with auto-memory: the model now automatically remembers what it learns across sessions, without you having to write anything down. Your project context, debugging patterns, and preferred approaches are remembered and recalled in the next session.
Projects in Cowork
The most recent addition: separate workspaces with their own files, instructions, and memory. Where previously you started a new session for every task, you can now maintain an ongoing project. Comparable to how OpenClaw organizes workspaces with long-term memory.
Connectors via MCP
OpenClaw integrates with more than fifty services. Anthropic answered this with a growing network of MCP connectors: Gmail, Google Drive, Google Calendar, Slack, Notion, Microsoft 365, Salesforce, DocuSign, and more. The count now stands at over 38 official connectors, and other MCPs can also be connected. Through both platforms, I can order groceries at Picnic or run an analysis of my historical purchases.
The comparison: where they overlap and where they differ
Security. The biggest difference. OpenClaw gives the AI full system access — very powerful but risky. Cisco called it a "security nightmare," and several critical vulnerabilities have been discovered. Claude Cowork runs in a sandbox: a shielded environment where the AI can only access the folders you explicitly share.
Cost. OpenClaw is free and open source, but you pay per API call to whichever AI model you use — expect around €5 to €20 per month with normal usage. Claude Cowork is included with a Pro subscription (€20/month) or Max subscription (€100/month). With intensive use, OpenClaw can actually turn out to be more expensive.
Ease of installation. Claude wins here by a mile. Download the app, pair your phone via QR code, done. OpenClaw requires a terminal, Node.js, API keys, and manual configuration of messaging channels. Not a problem for developers; a dealbreaker for marketers.
Flexibility. OpenClaw wins here, though the gap is shrinking. You choose which AI model to use (Claude, GPT, Gemini, or local models) and have access to more than 700 community skills. Until recently, Claude worked exclusively through its own app, but with the recent launch of Channels (Telegram, Discord), Anthropic is beginning to think multi-channel as well.
Scheduled tasks. Both platforms support recurring tasks. With OpenClaw, you configure cron jobs in YAML. With Claude, you type what you want in plain language and choose a frequency — no technical knowledge required. Moreover, Claude's scheduled tasks can now run in the cloud, meaning your computer doesn't even need to be on. With OpenClaw, your machine must always be running.
Integrations. OpenClaw offers more than fifty integrations and webhooks for external triggers. Claude Cowork now counts over 38 MCP connectors (Gmail, Slack, Drive, Notion, Microsoft 365, Salesforce) and is growing rapidly, but still lacks the webhook functionality that makes OpenClaw so flexible.
Remote control. The heart of the comparison. With OpenClaw, you send a WhatsApp message and the AI executes it on your computer, including full screen control. With Claude, you now have two options: Dispatch via the Claude app, or Channels via Telegram and Discord. Same idea, but Claude now also offers the chat apps that made OpenClaw big. And with computer use, Claude can now control your screen too: clicking, scrolling, and operating applications. The difference lies in the approach: OpenClaw provides full system access; Claude does it within safeguards, with explicit permission and automatic prompt injection detection.
What does this mean?
The conclusion is not that Anthropic copied OpenClaw. Cowork was already in development before Clawdbot went viral, and much of the underlying technology — such as the Model Context Protocol — was announced back in 2024. But the speed at which Anthropic launched features that answer the OpenClaw proposition point by point is striking.
Look at it from Anthropic's perspective: they saw a developer using their model to build the fastest-growing open-source project ever. Their first response was to protect the brand name. Their second response was to shut down OAuth. Their third response — and the only one that truly matters for most people — was to build a better, safer, more accessible alternative. With computer use, the last major gap has been closed: the full OpenClaw proposition is now available in Claude Desktop.
This is how a shared vision of what an AI assistant should be is emerging at breakneck speed: an AI that doesn't wait for you to ask a question but proactively executes tasks. That runs on your own computer, knows your files, uses your tools, and is reachable via your phone. A digital colleague, not a chatbot.
At the same time, they're also very different. OpenClaw is like the hobbyist chef's knife — incredibly sharp, fully customizable, but you need to know what you're doing (and the security risks are real). Claude (Cowork) is the professional kitchen appliance: less flexible, but working out of the box for everyone, with a sandbox that prevents your AI from uploading your entire hard drive to the internet.
What does this mean for marketing and communications professionals?
The features we've described aren't only relevant for developers automating scripts. They can also influence the daily work of marketers and communications professionals. A few examples:
Automate daily briefings. Set up a scheduled task that summarizes your inbox, Slack messages, and calendar every morning. No more scrolling for half an hour before you know what's going on.
Automatic reports. Weekly social media summaries, monthly content overviews, or quarterly reports that are automatically generated from your Google Drive or spreadsheets.
Delegate research while on the go. Via Dispatch, you can start a competitive analysis from the train that's ready when you arrive at the office.
Plugins for your specific role. The marketing plugin bundles capabilities for campaign analysis, positioning, and content strategy. Ready to use immediately after installation — no prompt engineering required.
The barrier to putting this technology to productive use has dropped dramatically. You don't need to open a terminal, write YAML, or configure API keys. You just need to describe what you want to do.
The lesson
We're going to see the OpenClaw story play out more and more in the coming years. A solo builder or small group of people proves there's demand for a certain type of AI tool. Big tech companies absorb that innovation by building comparable features with more usability and better security. And the real winners are us as users, because competition drives everyone toward better products.
What was a reason to install OpenClaw last month, you can now set up with two clicks in Claude Desktop. That makes it all the more important not to wait until your industry is "next in line," but to start learning to work with what's possible right now. The professionals who are already having Claude generate a morning briefing today will be light-years ahead in six months of colleagues who didn't get a perfect answer from ChatGPT two years ago and shelved it.
