Anthropic shipped Claude Code Channels today, letting you control a Claude Code session from Telegram. I set it up this afternoon. The pitch is irresistible: DM your bot from your phone, Claude executes on your Mac. MacStories built an entire iOS project wirelessly and the demo is genuinely impressive.

Then you start using it for real.

Messages get silently dropped. The bot shows "typing..." and nothing arrives. Close your terminal and the bot dies, messages lost permanently, no queue. Need to approve a permission prompt? Walk to your Mac. A version upgrade broke group messages with zero error output. This feels familiar for anyone who's watched Claude become load-bearing infrastructure and then buckle.

Getting here took more effort than it should have. The setup is rough, the documentation assumes you'll figure things out, and the failure modes are silent enough to make you question whether anything is happening at all. But once everything is configured and the pairing is locked down, it genuinely works. I've been running skills, deploying blog posts, and downloading media from my phone all evening. The gap between the concept and the execution is real, but the concept wins in the end.

Sources: