Anthropic released Channels — message AI from Telegram

Anthropic just shipped one of the most practical AI automation features of the year: Channels for Claude Code. You can now send a message to your AI coding agent from Telegram or Discord — and it messages you back. While Claude Code works on your project in the terminal, you control it from your phone.

The feature launched in Claude Code v2.1.80 as a research preview. It's already the #1 story on Hacker News — and for good reason. This turns a terminal-only AI agent into something you can supervise from anywhere.

How Channels actually works

A Channel is a bridge between an external messaging platform and your Claude Code session. Think of it like giving your AI assistant a phone number. Here's what happens:

This isn't just one-way alerts. It's two-way conversation. You message "hey, what files changed today?" from Telegram, and Claude checks your git history and sends the answer right back to Telegram. You say "run the tests" — Claude runs them and reports the results to your chat.

Set up Telegram in 5 minutes

The Telegram setup is the most popular option. Here's the full process:

Step 1: Create a bot in Telegram by messaging @BotFather and sending /newbot. Copy the token it gives you.

Step 2: Install the plugin and configure it:

/plugin install telegram@claude-plugins-official
/telegram:configure YOUR_BOT_TOKEN

Step 3: Restart Claude Code with channels enabled:

claude --channels plugin:telegram@claude-plugins-official

Step 4: DM your bot on Telegram. It sends a pairing code. Enter it in Claude Code:

/telegram:access pair YOUR_CODE
/telegram:access policy allowlist

That's it. Your AI agent now has a Telegram inbox.

Discord works the same way

If your team lives in Discord, the setup is nearly identical. Create a bot in the Discord Developer Portal, enable Message Content Intent, install the plugin, and pair your account. The bot needs permissions to view channels, send messages, read history, and attach files.

/plugin install discord@claude-plugins-official
/discord:configure YOUR_BOT_TOKEN
claude --channels plugin:discord@claude-plugins-official

What Claude can do through a Channel

Once connected, Claude has three tools available through the chat:

Photos sent to the bot are automatically downloaded and made available to Claude. Claude shows a typing indicator while it processes your request — just like chatting with a real person.

Security: only you can talk to your agent

Anthropic built this with a security-first approach. Every channel uses a sender allowlist — only Telegram or Discord accounts you've explicitly paired can send messages. Everyone else is silently dropped.

The pairing flow works like Bluetooth: you message the bot, get a one-time code, and approve it in your terminal. After that, only your account ID is whitelisted. For teams, organization admins must explicitly enable channels in their admin settings before anyone can use them.

Build custom channels for CI, monitoring, or webhooks

Telegram and Discord are just the starting plugins. Anthropic published a full reference for building your own channels using the MCP protocol (the standard that connects AI tools together).

The possibilities are immediately practical:

• CI/CD alerts: Your build fails → Claude gets a webhook → it reads the logs, finds the bug, and starts fixing it
• Monitoring: Server alert fires → Claude investigates and reports back
• Team chat: A Slack channel for your AI assistant (community plugin coming)

A minimal webhook receiver is just ~30 lines of TypeScript:

import { Server } from '@modelcontextprotocol/sdk/server/index.js'
import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js'

const mcp = new Server(
  { name: 'webhook', version: '0.0.1' },
  { capabilities: { experimental: { 'claude/channel': {} } } }
)
await mcp.connect(new StdioServerTransport())

Bun.serve({
  port: 8788,
  hostname: '127.0.0.1',
  async fetch(req) {
    const body = await req.text()
    await mcp.notification({
      method: 'notifications/claude/channel',
      params: { content: body }
    })
    return new Response('ok')
  }
})

Who should try this — and who should wait

Developers who leave Claude Code running on long tasks — this is built for you. Instead of checking your terminal every 10 minutes, Claude pings you on Telegram when it needs input or finishes a task.

Teams managing multiple projects — set up a Discord bot per project. Your AI agents report status through Discord channels your team already watches.

Automation enthusiasts — the webhook channel means any system that can send HTTP can now talk to Claude. Connect your home automation, your CI pipeline, or your monitoring stack.

Note: Channels are in research preview. You need Claude Code v2.1.80+, a claude.ai login (API keys don't work yet), and Bun installed. Only Anthropic-approved plugins are allowed during preview — custom channels require a development flag.

The bigger picture: AI agents that never go offline

Channels are part of a broader push at Anthropic to make Claude Code an always-available assistant. Combined with Remote Control (for driving sessions from your phone) and scheduled tasks (for timer-based polling), Claude Code is evolving from a terminal tool into something closer to a persistent AI coworker.

The 16 official plugins already available — including GitHub, Slack, Linear, Stripe, Firebase, and Playwright — hint at where this is heading. Each one can potentially become a channel. Your AI agent doesn't just write code anymore. It listens.

Related Content — Get Started with Easy Claude Code | Free Learning Guides | More AI News