What Are Claude Code Channels?
Anthropic has officially launched Claude Code Channels, a new feature that transforms how developers interact with AI coding assistants by enabling two-way communication through Telegram and Discord. Released on March 20, 2026, as a research preview, this update allows developers to message their Claude Code sessions directly from their phones — and get full responses back — without ever opening a terminal.
The feature represents a fundamental shift from the traditional synchronous “ask-and-wait” model to an asynchronous, always-on AI coding partnership. Instead of being tethered to a desktop terminal, developers can now fire off coding instructions, check on project status, or debug issues from anywhere using messaging apps they already use daily.
How Claude Code Channels Work Under the Hood
At its core, a Claude Code Channel is an MCP (Model Context Protocol) server that pushes events into a running Claude Code session. When a developer sends a message via Telegram or Discord, that message arrives in the active Claude Code session on their local machine, where Claude processes the request with full filesystem, git, and MCP tool access. The response then flows back through the same messaging platform.
The technical foundation relies on the Model Context Protocol, the open-source standard Anthropic introduced in 2024. MCP acts as a universal connector for AI systems, providing a standardized way for AI models to interact with external data and tools. Claude Code Channels extend this protocol to support real-time, bidirectional messaging between developers and their AI coding assistant.
Setting up a channel requires Claude Code v2.1.80 or later and a claude.ai account on the Pro, Max, or Enterprise tier. Developers create a bot through Telegram’s BotFather or the Discord Developer Portal, install the corresponding plugin in Claude Code, configure their bot token, and restart with the --channels flag enabled. A pairing process ensures only authorized users can push messages to the session.
Security and Enterprise Controls
Anthropic has built security into the core of Claude Code Channels. Every approved channel plugin maintains a sender allowlist, meaning only verified user IDs can push messages — everyone else is silently dropped. The pairing process requires a unique code that must be confirmed in both the messaging app and the Claude Code terminal.
For enterprise customers, channels are disabled by default and must be explicitly enabled by an organization admin through the claude.ai admin settings panel. This gives IT teams full control over whether their developers can use the feature, addressing common concerns about data security and unauthorized tool access in corporate environments.
Additionally, being configured in the MCP settings file alone is not enough to activate a channel. Servers must be explicitly named in the --channels startup flag, adding an extra layer of intentional activation that prevents accidental exposure.
Why This Matters: The OpenClaw Competition
The timing of Claude Code Channels is widely seen as a strategic response to the growing open-source AI coding movement, particularly OpenClaw, which has been gaining traction among developers who want flexible, self-hosted AI coding solutions. Multiple industry publications have described the feature as Anthropic’s answer to these competitive pressures.
While OpenClaw and similar projects typically require dedicated hardware, complex self-hosting setups, or third-party bridges to achieve similar functionality, Anthropic’s approach leverages its existing Claude Code infrastructure and plugin architecture. Developers get always-on AI coding assistance through apps they already have installed, with no additional servers to maintain.
The plugin-based architecture also means more platforms can follow beyond Telegram and Discord. Anthropic has published an open channel reference specification, allowing developers to build custom channels for systems like Slack webhooks, CI/CD pipelines, error trackers, and deployment monitoring tools.
Beyond Chat: Webhooks and Automation
Claude Code Channels are not limited to human-to-AI chat. The feature also supports webhook receivers, enabling automated systems to push events directly into a Claude Code session. This means a failed CI build, an error tracker alert, or a deployment pipeline notification can arrive in a session where Claude already has the developer’s files open and context about what they were working on.
This positions Claude Code as more than just a reactive coding assistant. With channels active, it becomes a proactive development partner that can respond to external events, triage alerts, and begin investigating issues before the developer even notices something went wrong.
Availability and What’s Next
Claude Code Channels are currently rolling out as a research preview, with Telegram and Discord support available as the initial launch platforms. The --channels flag syntax and protocol specifications may evolve based on developer feedback during the preview period.
Pro and Max individual users can start using channels immediately by opting in per session, while Team and Enterprise organizations need admin approval first. Anthropic has indicated that the feature will expand based on community feedback, with custom channel development already possible through the published reference documentation.
The launch adds to an already packed March 2026 for Anthropic, which also announced a $100 million investment in the Claude Partner Network, the launch of The Anthropic Institute, and a new Asia-Pacific office in Sydney. With Claude Code Channels, the company is making a clear bet that the future of AI-assisted development is not just smarter models, but smarter ways to stay connected to them.