Anthropic Unveils 'Channels' to Control Claude Code Sessions Directly from Mobile

Anthropic has introduced Channels, a new feature that makes Claude Code feel more natural to use on mobile. With this update, users can give instructions to the AI coding agent from their smartphones via messaging apps like Telegram or Discord—as easily as texting a friend.

Channels currently supports Telegram and Discord first. This means users no longer have to sit in front of a desktop development environment; they can connect to an ongoing Claude Code session and issue commands while on the go or away from their desk. For example, they can quickly send requests from mobile like “Check the part that just errored again,” “Summarize the test results,” or “Continue refactoring that code.”

What makes this feature especially interesting is how development tools are increasingly moving into the communication environments people already know. Traditionally, using a coding agent meant working inside conventional development tools like an IDE or terminal, but now messaging apps themselves are becoming agent control interfaces. Instead of learning a new tool, users can collaborate with AI naturally within apps they already use every day.

Another point worth noting is the continuity of work context. It goes beyond simply sending commands from mobile—users can pick up where their existing session left off and keep the work going, which makes it highly practical. Developers can stay on top of their in-progress tasks and deliver instructions in real time even during meetings or while out of the office. This also aligns well with the growing importance of remote collaboration and asynchronous work.

In the end, Channels is an update that lets people use AI coding agents more often, more casually, and as part of everyday life. It marks a meaningful shift away from development environments tied to a specific place and device, toward experiences that continue anytime, anywhere. If support for more platforms and richer features follow, messenger-based AI development workflows could become mainstream faster than we expect.