Anthropic lets you text Claude Code from Telegram and Discord

Picture this: your CI pipeline fails while you are standing in line for a coffee. Usually, that means rushing back to your laptop, finding Wi-Fi, and digging through terminal logs.

Anthropic wants you to just pull out your phone instead.

Anthropic quietly rolled out a preview of Channels for Claude Code, letting developers control running coding sessions via Telegram and Discord. Claude Code engineer Thariq Shihipar confirmed the launch on X.

Texting your terminal

Channels is built on Anthropic's Model Context Protocol (MCP), acting as a two-way bridge between your local development environment and your messaging apps.

External triggers like webhook callbacks or CI build failures can now feed directly into Claude Code. The AI analyzes the error, pings your chat with a diagnosis, and waits for your input.

From there, you just text back a instruction. Claude Code runs the fix script on your machine in the background and reports back the outcome.

Setup is lightweight and resembles spinning up a standard bot. You grab a token, bind credentials, launch it with flags, and you are up in minutes.

That said, it is very much an early preview. Your session still depends on a persistent local process via tmux or screen, and it currently lacks robust API key authentication.

Hands-off agents vs. deep co-pilots

The move highlights a split in how tech giants approach AI software engineering.

One camp, led by tools like OpenClaw alongside pushes from Nvidia NemoClaw and OpenAI, aims for full hands-off autonomy. Their goal is minimizing human touch through end-to-end task execution.

Anthropic is choosing human-in-the-loop collaboration. Instead of handing over the wheel, Channels keeps the developer in control while untethering them from the physical desk.

Why it matters

For developers, this turns emergency on-call duties into an asynchronous mobile task. You do not need an entire mobile IDE just to approve a hotfix or triage a broken test.

However, it is still an experimental setup. If every tiny terminal action requires constant manual replies over chat, the back-and-forth friction could easily outweigh the convenience.

For the broader landscape, Anthropic is proving that developer infrastructure does not always need reinvention. Leveraging MCP to pipe agentic tools into everyday apps is a remarkably clever shortcut.

We think this pragmatic approach makes a lot of sense. Before autonomous agents can be trusted with production systems, keeping humans seamlessly looped in via their phones is the safest bridge forward.