Anthropic has released Claude Code Channels as a research preview, enabling developers to send messages from Telegram or Discord directly into a running Claude Code session on their local machine. The session processes requests using full filesystem access, MCP (Model Context Protocol) servers, and git integration, then replies through the same messaging app. This represents a fundamental shift from synchronous terminal interaction to asynchronous, mobile-friendly AI agent control. What Makes Claude Code Channels Different from Traditional AI Integrations? Unlike integrations that spawn fresh cloud sessions or poll for updates, channel messages arrive in the session you already have open with your real files and project context. When Claude replies through a channel, the actual response appears on Telegram or Discord, while your terminal shows the tool call and confirmation. This means a developer can be away from their desk, send a message from their phone asking Claude to check the status of a failing test, and Claude will execute the work locally against the actual codebase and reply in the same chat thread. The timing of Claude Code Channels is significant. OpenClaw, an open-source agentic AI platform, gained massive traction in March 2026 by enabling developers to manage fleets of AI agents through messaging apps like WhatsApp and Telegram from their home computers. VentureBeat described Claude Code Channels as an "OpenClaw killer" that brings similar mobile agent control capabilities into Anthropic's ecosystem. While OpenClaw offers a fully open-source, multi-model approach, Claude Code Channels integrates directly with Anthropic's models and the full Claude Code development environment. How to Set Up Claude Code Channels for Your Workflow - Create a Bot: Generate a bot token through BotFather for Telegram or the Discord Developer Portal for Discord, then provide those credentials to Claude Code. - Install the Plugin: Use the /plugin install command to add the Telegram or Discord channel plugin to your Claude Code environment. - Enable and Pair: Restart Claude Code with the --channels flag, send a message to your bot to receive a pairing code, and enter that code in your Claude Code session to establish the connection. - Configure Allowlists: Verify that only trusted contacts are added to your sender allowlist, since only explicitly paired and approved IDs can push messages into your session. The feature requires Claude Code v2.1.80 or later and is available to Pro and Max individual users by default. Team and Enterprise organizations must have an admin explicitly enable channels in managed settings before members can use them. How Does Security Work in Claude Code Channels? Every approved channel plugin maintains a sender allowlist. Only IDs that have been explicitly paired and approved can push messages into a session. All other senders are silently dropped. The pairing process requires physical access to both the messaging platform and the Claude Code terminal: a user sends a message to the bot, receives a pairing code, and enters that code in their Claude Code session. The allowlist also controls permission relay. If Claude encounters a permission prompt while the developer is away, channels with the relay capability can forward these prompts to the messaging app for remote approval or denial. Only allowlisted senders should be trusted with this authority. One important limitation is that channels only work while the Claude Code session is actively running. For always-on setups, users need to run Claude in a background process or persistent terminal. Events do not queue when the session is closed. What Makes the Plugin Architecture Significant? Channels use a plugin-based design that makes the system extensible. Telegram and Discord are the two officially supported platforms in the research preview, but the architecture is designed so additional platforms can follow. Each channel is installed as a plugin via the /plugin install command and configured with platform-specific credentials. The community has already begun requesting Slack, WhatsApp, and iMessage support, and Anthropic has published the full plugin source code on GitHub under the claude-plugins-official repository for developers who want to build custom channels. The plugin architecture is perhaps more significant than the initial Telegram and Discord support. By publishing source code and providing a channels reference for custom builds, Anthropic is positioning Claude Code Channels as a platform rather than a feature. If the community builds plugins for Slack, Microsoft Teams, and other enterprise communication tools, channels could become the primary interface for remote AI agent management. Claude Code Channels fills a genuine gap in the AI coding agent workflow by enabling mobile-first interaction with running Claude Code sessions. The enterprise controls suggest Anthropic is already thinking about corporate adoption. In organizations where developers use Claude Code for production work, the ability to trigger and monitor agent tasks from a mobile device could significantly increase developer productivity during off-hours incidents and collaborative workflows.
