01What Claude Code Channels Is and How It Works
Claude Code Channels is a new Anthropic feature that lets you message your Claude Code AI agent from your phone through Telegram or Discord. It was released as a research preview in March 2026. This means you can talk to your coding agent without sitting at your computer or opening a terminal.
The feature works through MCP servers, which are the technical bridge that connects Claude Code to your chat apps. Once set up, your AI agent appears in the chat window just like a co-worker you message. You can send tasks, check on progress, or ask questions from anywhere.
This is a big shift because Claude Code used to live only in your terminal. Now it can be with you wherever you carry your phone. Anthropic says this is focused on developers who want something hackable, meaning you can customize it and build on top of it.
02Key Features and What's New
Claude Code Channels adds two major ways to reach your AI agent: Telegram and Discord. You can send messages from your phone and the agent will respond. This is the main new feature, but it's part of a bigger wave of updates in March 2026.
- Voice mode in 20 languages, so you can talk to Claude Code instead of typing.
- 1 million token context window for handling very large codebases or projects.
- Forward slash loop command for recurring tasks, like checking deploy status every five minutes.
- Remote control to continue a coding session from your phone, tablet, or web browser.
- Opus 4.6 as the default model now, with the 'ultra think' keyword for maximum reasoning.
- Forward slash effort command to adjust how hard the AI thinks about a problem.
These features work together to make Claude Code more like an always-on assistant, not just a terminal tool. You can start a task at your desk, then leave and monitor it from your phone.
03How to Set Up Claude Code Channels
Setting up Claude Code Channels is straightforward, and you can even use Claude Code itself to help you. The official guide is live on GitHub. Follow these steps to get started:
- Update Claude Code on your computer by running the NPM update command.
- Choose which channel you want to use: Telegram or Discord.
- Follow the setup guide on GitHub to create a bot for your chosen channel.
- Connect the bot to Claude Code using MCP servers, which is the technical bridge.
- Test the connection by sending a message from your phone to your bot.
- Start sending tasks and commands to your AI agent from anywhere.
You can also ask Claude Code to help you with the setup. Just describe what you want to do, and it can guide you through the process. This makes it easier even if you're not a deep technical expert, though it helps to be somewhat technical.
04How to Use It and Real Examples
Once channels are set up, using Claude Code is like texting a co-worker. You open Telegram or Discord, find your bot, and send a message. For example, you could type, 'Check the deploy status every five minutes' and Claude Code will do that using the loop command.
You can also use it for everyday tasks. Imagine you're on vacation and want to monitor your systems. You can send a message like 'Any issues with the app?' and get a report.
Or you could ask it to prepare a summary of the latest AI news every hour.
Anthropic's product manager, Tariq, explained that this is about giving people different options to talk to Claude remotely. It's designed for developers who want something hackable, meaning you can connect it to your own workflows and tools. This is not a locked-down product; it's pipes you can build with.
The key is to treat it like a team member. Set tasks, ask questions, and let it run in the background. You can even use it to check on long-running coding tasks while you're away from your desk.
05Results and Comparisons with OpenClaw
Anthropic's move directly targets OpenClaw's biggest advantage: access. OpenClaw, the most starred software project on GitHub with over 300,000 stars, lets you message your AI through WhatsApp, Telegram, Discord, iMessage, and 30+ other platforms. It also runs 24/7 and has persistent memory.
Claude Code previously lacked these features.
However, Claude Code has been closing the gap. In March 2026 alone, it added voice mode, a 1 million token context window, loop commands, and remote control. Now channels add the messaging capability.
This is a direct response to OpenClaw's popularity.
Security is a major differentiator. Researchers found 135,000 exposed OpenClaw machines and 50,000 vulnerable instances that could be taken over remotely. A security audit of OpenClaw's skill registry found about 12% of skills (341 out of 2,857) were malicious.
In contrast, Anthropic offers security reviews, enterprise support, and a company you can hold accountable.
Claude Code also keeps humans in the loop. It asks for permission before implementing commands and sandboxes operations. OpenClaw acts autonomously, which can be magical but also risky.
A Meta executive even reported her OpenClaw deleted her entire email.
06Tips, Limits, and Mistakes to Avoid
When using Claude Code Channels, remember that it's still in research preview. That means it might have bugs or limitations. Don't rely on it for critical tasks without testing first.
- Keep security in mind: Even though Anthropic is safer than OpenClaw, you're still giving an AI access to your code and messages. Use strong authentication and don't share sensitive data unless necessary.
- Learn the loop command: It's powerful for scheduling recurring checks, but set realistic intervals to avoid charging up your API usage.
- Don't expect autonomous behavior: Claude Code asks for permission before major actions. This is intentional, so don't try to force it to act without oversight.
- Update regularly: New features are being added fast. Running the NPM update regularly keeps you current.
- Understand the limits: Claude Code doesn't have OpenClaw's persistent memory across sessions. It resets memory between sessions, so you may need to remind it of context.
For OpenClaw users, be aware of the security risks. Many community plugins and skills are malicious. Only install from trusted sources and review code when possible.
07Who It's For and Why It Matters
Claude Code Channels is for developers and technical users who want to manage coding projects from anywhere. But it also matters for non-developers. The tools developers use today become the foundation for products regular people use tomorrow.
For example, the ability to run background tasks and communicate over Telegram and Discord will flow into automation tools, business software, and personal assistants. AI agents are already writing about 41% of all code, and that number is climbing. Soon, many professionals will message AI agents like co-workers on Slack or text.
If you're a business owner, freelancer, or creator, learning to use these tools is not optional anymore. It's a skill that separates people who thrive from those who struggle. You don't need to learn to code, but you do need to understand what AI agents can do.
Julian Goldie, the host, emphasizes that the gap between AI users and non-users is widening every week. Starting now gives you a competitive advantage that will be hard to close later.
08The Broader War Between Anthropic and OpenClaw
This announcement is part of a larger power struggle. OpenClaw, backed by OpenAI, has been the most viral open-source project ever, reaching 300,000 GitHub stars in 60 days. It runs on your phone, has persistent memory, and works with over 30 platforms.
But it also has serious security flaws.
Anthropic is responding with Claude Code's new features. They're not just adding access; they're adding it with safety. Claude Code's approach is to keep humans in the loop, while OpenClaw's philosophy is full autonomy.
Both have value, but for most people, safety is crucial.
OpenClaw is transitioning to an open-source foundation with OpenAI backing after its creator, Steinberger, joined OpenAI. This means OpenClaw will continue to grow and compete. The AI world in 2026 is messy, with companies funding each other's competitors.
But Anthropic has billions in revenue, security teams, and backing from Amazon ($4 billion) and Google. The race for the access layer is not even close in terms of resources. Yet OpenClaw still has a lead in day-to-day automation and being completely free with local models.
09What You Should Do Next
If you're a developer, the first step is to update Claude Code and set up channels. You can use the NPM update and follow the GitHub guide. Don't miss out on this new capability.
If you're a non-developer, start learning how to use AI tools, not how to code. Understand what they're capable of so you can design your workflows around them. Look for repetitive tasks in your job that AI could handle, like follow-ups, spreadsheets, or scheduling.
For managers, think about where your team spends time on tasks that AI could take over. Free up your people for work that requires human judgment. The companies that figure this out now will have a huge advantage.
Julian Goldie suggests joining his AI Profit Boardroom to learn practical AI workflows. But even without that, start experimenting with tools like Claude Code. The pace of change is fast, and the timing to get ahead is now.
