Anthropic Shipped Claude Channels. Your AI Agent Can Now Text You Back.
Claude Channels quietly changes the relationship between people and AI agents. Here's what it actually means if you run a business.
Until very recently, every interaction with an AI agent had the same shape. You sit down. You open the tool. You give it a task. You wait. You check. You iterate. Every cycle requires your presence. Walk away and the session stalls, the output piles up unseen, or a permission prompt freezes everything until you come back.
That constraint just changed.
On March 20, 2026, Anthropic shipped a feature called Claude Code Channels. It lets Claude’s agentic tool communicate with you through Telegram, Discord, and iMessage. You send a task from your phone. Claude does the work on your computer. When it’s done — or when it needs you — it messages you back.
Most of the coverage has been aimed at developers. That’s where it launched. But the real audience is anyone who runs a business and uses AI in any serious way. Because Channels doesn’t just change how you interact with Claude. It changes the fundamental model of what AI agents are.
— AI Maker newsletter, Q1 2026 review of Claude updates
What changed and why it matters
Before Channels, Claude Code was a terminal tool. Powerful, capable, genuinely impressive — but you had to be present. It was a tool you operated. The moment you stepped away, it stopped being useful.
Channels changes the direction of information flow. Instead of you going to Claude, Claude comes to you.
Technically, a Channel is an MCP (Model Context Protocol) server — a bridge that connects a running Claude Code session to an external messaging platform. Forget the acronym. The practical picture: you create a bot in Telegram or Discord, link it to your Claude session, and from that point on you have a two-way conversation with your agent from any device, anywhere.
Your files stay on your machine. Your codebase stays local. Only the text of the conversation passes through the external platform. You’re not pushing sensitive data to a third-party server. You’re just giving Claude a phone number.
You run a small product company. Monday morning means a weekly status report: sales numbers, support ticket volume, any anomalies in the data. You set Claude up with access to your data folder and start a Channels session before you leave for the gym. By the time you’re back, the report is waiting in your Telegram. You didn’t ask for it that morning. It just ran.
From tool to worker
This is the frame that matters for business people.
Calculators wait for you. Assistants don’t. For years, AI tools have been calculators — you bring the input, you collect the output, you close the window. Channels is the first step toward AI that behaves like the second kind.
The GLN-7.5 technology blog put it plainly: Channels, along with Dispatch, Scheduled Tasks, and Cowork, represent a shift from AI as a tool you use to AI as a worker that handles things. All four features launched in March 2026. Channels is the messaging layer — the way you stay connected to work that’s running without you in the room.
The AI Maker newsletter, which tested every Claude update across Q1 2026, described the behavioral shift in three words: the output finds you. That’s the change. Not smarter models. Not bigger context windows. Just the simple inversion of who initiates contact.
You’re managing a newsletter and a product launch simultaneously. You ask Claude to draft three subject line options for next week’s send based on the draft sitting in your content folder. You send the message from your phone while in a meeting. Ten minutes later, three options are in your Discord. You pick one, reply with your choice, Claude updates the draft. You haven’t opened a laptop.
The OpenClaw backstory
If you’ve been reading CO/AI, you know the claw-code story. Two engineers built the fastest-growing repo in GitHub history on airplane Wi-Fi — by texting their AI agents, not sitting at terminals. They weren’t exceptional in raw talent. They were exceptional in how they’d set up their systems.
OpenClaw, the open-source AI agent framework with 250,000 GitHub stars, had already figured this out. Remote control of AI agents through Telegram and WhatsApp wasn’t new — the developer community had built it themselves, with workarounds, custom infrastructure, and a tolerance for rough edges.
Channels is Anthropic’s answer. Official plugins. Audited code. A security model built in from the start. VentureBeat called it an “OpenClaw killer.” The more honest framing: it takes something the developer underground had been hacking together for months and makes it accessible without a custom server setup.
For business people, that distinction matters. You don’t need to be Bellman running five Codex Pro subscriptions simultaneously to use this. The Shareuhack technical review put setup time at roughly 15 minutes. Start a bot in Telegram via BotFather, run a configure command inside Claude Code, pair your account with a code, done.
Your team already uses Discord. An engineer kicks off a long refactor before end of day and connects their Claude session to the team’s #ai-work channel. When Claude hits a decision point, it posts a question. A senior engineer replies from their phone during the commute home. The session continues. Nobody stayed late. Nobody missed the window.
What you need to know before you set it up
There are real constraints worth understanding before you build a workflow around this.
You need a paid account. Pro at $20/month or above. Free plan users can’t use Channels.
Your machine needs to stay on. Channels only works while a Claude Code session is actively running. If you close your laptop, incoming messages go unanswered. For always-on behavior you need to keep a session running on a persistent machine — a dedicated desktop, a server, a cloud VM. This is the biggest practical limitation for small teams. The FindSkill.ai guide on Telegram setup called it the constraint most tutorials skip. It’s real.
It’s a research preview. The commands and setup flow may change before general availability. Build habits around it, but build them loosely.
Security is your responsibility. The three-layer security model covers who can message Claude. It doesn’t automatically limit what Claude does after receiving a message. The Shareuhack technical breakdown described this as a risk dial, not an off switch — standard mode for full protection, skip-permissions mode for unattended work with a more restricted scope. Treat it like any access-control decision: think about what you’re exposing before you enable it.
Where this goes
Channels is a research preview of something much larger. Cowork, Anthropic’s fully cloud-based agent mode, already runs tasks completely independently — no open session required, no local machine. Scheduled Tasks automates recurring work without any human trigger. Dispatch lets you start new work from your phone without touching your computer.
Taken together, the picture is clear. AI agents that require your constant presence are giving way to agents that work on your behalf and report back. The question for most business people isn’t whether to engage with this shift. It’s whether to engage with it now or wait until the tools are more polished and the gap between early adopters and everyone else is wider.
The signal in the claw-code story wasn’t the GitHub star count. It was the engineers who built it. They weren’t at their terminals. They were on planes, texting their agents. The infrastructure to do that, built informally in the open-source community, is now shipping as a first-party Anthropic feature.
You have a phone. Claude has a bot. Set it up.
Anthropic
AI Agents
Automation
OpenClaw
Telegram
Discord
Recent Blog Posts
What Did You Do Today?
There's a saying in Jackson Hole. You hear it at the coffee shop on the square, on the chairlift at the Village, in the bars after a day on the mountain. It goes like this: It's not what you do. It's what you did today. I've been thinking about that line all weekend. Because Sam Lessin dropped a piece arguing that AI isn't just a labor crisis — it's a meaning crisis. And Goldman Sachs just published 40 years of data proving that when technology displaces workers, the damage doesn't heal. It scars. Ten percent slower earnings growth for the...
Apr 3, 2026Claw-code Broke GitHub’s Star Record in 24 Hours. Two Engineers Did It on an Airplane. Here’s What That Means for Your Business.
Here's the number: 100,000. That's how many GitHub stars a repository called claw-code collected in roughly 24 hours. Not a year. Not a month. One day. By the time a live stream was done discussing it, the counter was climbing by a thousand stars every ten minutes. Nobody in the room could remember seeing anything grow that fast. Because nothing had. I watched it happen in real time. I'd met the two engineers behind it the weekend before at an AI hackathon in San Francisco. Within 72 hours of shaking hands, they'd built the fastest-growing repo in GitHub history —...
Mar 15, 2026Elon Musk Doesn’t Run Six Companies. He Runs One Router.
On Wednesday morning, Andrej Karpathy — the man who taught a generation of engineers to build neural networks — told everyone to stop writing code. Manage the agents that write it, he said. The guy who wrote the playbook just rewrote it. We covered the implications in our Signal/Noise briefing: who builds the Cisco for agents, what happens when agents get wallets, why the Fastenal vending machine is the best metaphor for the AI economy. But the more we pulled on the thread, the more a different question emerged. Not about code. Not about models. About organizations. Specifically: what does a...