Claude Code Channels: Anthropic's Always-On Agent Is Now in Your Pocket
Agentic AIAnthropic’s Claude Code can now receive messages from Telegram and Discord, respond to CI results, and act on webhook notifications — all without a developer sitting at the terminal. The “channels” feature, shipped March 20 as a research preview, is the first native implementation of a workflow that third-party tools had been patching together for months.
Why This Changes the Agent Equation
The distinction between a coding assistant and a coding agent has always come down to one question: does it wait for you, or does it work while you’re away? Claude Code has consistently been one of the most capable tools in the former category. Channels moves it into the latter.
The practical shift is significant for developers running any kind of automated pipeline — CI/CD systems, monitoring stacks, deployment workflows. Rather than returning to the terminal to assess an overnight test run or a staging environment alert, the agent is already aware and can have acted on it. For teams, this closes the loop between automated systems and AI reasoning without requiring a human intermediary.
How It Works
Channels run through MCP servers — the Model Context Protocol that Anthropic open-sourced and donated to the Agentic AI Foundation under the Linux Foundation in December 2025. The architecture is bidirectional: external events push into a running Claude Code session, and Claude responds through the same channel.
The research preview currently supports Telegram and Discord. Developers can build custom channels for any platform that exposes a webhook or message API — Slack, PagerDuty, GitHub, or internal tooling. Requirements are narrow: Claude Code version 2.1.80 or later, a claude.ai login (API keys are not supported), and, for Teams and Enterprise organizations, explicit feature enablement.
The approach draws direct comparison to OpenClaw — the category of tool that let users control Claude Code remotely via chat messages. The difference is that Channels is a commercial implementation built on a standardized protocol, with Anthropic owning the integration surface directly rather than depending on third-party harnesses.
The So What
For solo developers and small operators, the immediate application is an agent that responds to production events at any hour — handling a failed deploy, investigating an alert, or continuing a long-running task — without requiring manual restart. The MCP foundation makes the integration surface extensible from day one: any channel you can build an MCP server for is a channel Claude Code can receive.
The research preview label signals roughness at the edges, but the underlying architecture is the same one now underpinning a growing range of agentic tooling across the industry. Teams evaluating whether to build internal agent infrastructure should treat this as the baseline they’re competing with.
Photo by César Gaviria on Pexels
Content created with AI assistance and reviewed for accuracy.
Join the conversation
Stack Insiders is our free community for readers who want to go deeper — share resources, ask questions, and connect with others across every vertical we cover.
Join Stack Insiders →