Anthropic has released Claude Code Channels, a feature that lets developers send messages to a running Claude Code session directly from Telegram or Discord. A developer can type a prompt on their phone, and Claude executes the task on their desktop or server and replies through the same messaging app.
The feature is built on the Model Context Protocol (MCP), an open standard Anthropic introduced in 2024. A channel acts as a two-way bridge: incoming messages are injected into the active Claude Code session as events, Claude processes the request and uses a reply tool to send the response back to the messaging platform.
One early tester, MacStories editor John Voorhees, built and deployed an iOS app from his iPhone through Telegram, compiled article lists, and processed podcast audio entirely from his phone while his Mac did the work. He noted that setup was slightly fiddly and required the Bun JavaScript runtime, which Anthropic acquired in December 2025.
A significant limitation remains. When Claude Code requires permission to run a file operation or shell command, the session pauses until someone physically approves it on the host machine. The only workaround disables all permission checks for that session.
Anthropic describes the feature as a research preview. Currently, only Telegram, Discord, and a local test interface called Fakechat are supported. Slack, WhatsApp, and other platforms are not yet available. The feature requires a claude.ai subscription. API key users cannot access it.
Enterprise and Team accounts have Channels disabled by default and require an administrator to enable it manually.
Sources: VentureBeat, Implicator.ai
