On March 20, Anthropic quietly launched a feature that has caught the attention of many developers - Claude Code Channels (Research Preview). Simply put: your Claude Code is still running on your local machine, but you're already on the subway, pull out your phone and send a Telegram message, and it can receive instructions, report progress, and continue working.
A member of the official team, Thariq, first posted a demonstration video on X. In the video, Claude Code was running npm tests, and a Discord user asked, "Is the build green yet?" Claude replied, "Still running tests – ~2 min. I'll ping you when it's done." - Then the Telegram side said, "Ship it when green 🚀," and the entire conversation felt as smooth as collaborating with a real engineer. The post caused a stir in the community.

How Does It Work
Channels are based on the MCP plugin mechanism, connecting external messaging platforms like Telegram and Discord to an ongoing Claude Code session, enabling bidirectional real-time communication. You can send commands, ask for progress, and transfer files; Claude can proactively push notifications, edit previous messages, and show a "typing" status - attachments are also supported, with Telegram allowing up to 50MB per file, and Discord supporting up to 10 files each at 25MB.
Currently, the research preview supports two official plugins: Telegram and Discord, plus a Fakechat demo plugin for local testing.
How to Configure
Prerequisites: Claude Code v2.1.80 or higher, logged in with a claude.ai account (not supporting direct API Key connection), and Bun installed locally.
The steps are straightforward:
- Install the plugin:
/plugin install telegram@claude-plugins-official, then/reload-plugins - Enter Bot Token:
/telegram:configure <your Token>(Token obtained from BotFather or Discord Developer Portal) - Start a session with Channels:
claude --channels plugin:telegram@claude-plugins-official - Secure pairing: Send any message to the Bot to get a 6-digit pairing code, then return to Claude and enter
/telegram:access pair <code>, then switch to whitelist mode to lock access permissions
Team/Enterprise users need administrators to enable channelsEnabled in the claude.ai backend first.
The full documentation is available at code.claude.com/docs/en/channels, and the READMEs for Telegram and Discord are in the official GitHub plugin repositories.
Some Things to Note
The session must remain active - this is a prerequisite; cloud servers or background persistence are acceptable. If Claude is waiting for permission confirmation, the session will pause; if you're in a hurry, you can add the --dangerously-skip-permissions flag, but this is only recommended for completely trusted environments. Currently, the plugin is limited to the official Anthropic repository, and syntax and protocols may be adjusted in the final version.
What the Community Thinks
Reactions were lively on X and Reddit. Some directly shouted, "OpenClaw is dead," while others compared their own self-built Telegram/Discord Bots, generally feeling that the official version is more deeply integrated and has a more complete security mechanism. Many also expressed hope for support for more platforms - iMessage and Slack were mentioned most often.
The significance of this feature goes beyond just "sending commands from your phone." Claude Code started as a local CLI tool, and previously added features like Remote Control and Loop scheduled tasks. Channels take it a big step further: making it truly an event-driven, remotely addressable asynchronous agent - you don't need to stay in front of your computer, and it keeps working.
It is currently available to Pro/Max users, and Anthropic says it will continue to expand supported platforms based on feedback. GitHub accepts issue submissions.
