CopilotKit has released Channels, an open-source SDK that brings AI agents into Slack, Microsoft Teams, and Discord with native, interactive interfaces. The tool allows agents to maintain their existing tools, models, and business logic while gaining a native presence in the communication platforms where teams already work.
According to the project documentation, Channels supports multiple agent frameworks. Developers can use CopilotKit’s built-in agent or connect LangGraph, CrewAI, Mastra, Pydantic AI, Google ADK, and other AG-UI-compatible agents without modifying the agent’s core logic.
The SDK handles several key capabilities: agents can understand conversations, stream responses, call tools, work with files, render interactive UI, and pause for human approval. Messages are rendered once and automatically converted to platform-specific UI—Slack Block Kit for Slack, Adaptive Cards for Microsoft Teams—ensuring native appearance on each platform.
Setup follows a streamlined path. Developers can run npx copilotkit@latest channels setup to install a channels-setup skill, which fetches current workflow documentation and guides agents through platform selection and framework choice. Alternatively, developers can install the Slack setup skill directly on disk and skip the hosted guide.
Channels operates through a managed connection model. The developer’s agent and application logic run in their own infrastructure, while CopilotKit Intelligence manages platform connections and delivers each conversation turn to the long-running Channels process. This separation keeps bot tokens and signing secrets under developer control in .env files.
The technical architecture requires Node.js 22 or later and involves installing @copilotkit/channels and @copilotkit/runtime packages. A long-running Node process or container handles the Channels listener, which receives messages from Slack or Teams via CopilotKit Intelligence, runs the agent, executes tools, and renders results back into the conversation.
On each turn, a person messages the app in Slack or Microsoft Teams, CopilotKit Intelligence receives the platform event and delivers it to the Channels process, Channels runs the agent and executes tools, and Intelligence sends native platform UI back into the conversation.
Managed connections for Slack and Microsoft Teams are available through CopilotKit Intelligence, with more channels coming. The SDK is MIT licensed and open source. CopilotKit Intelligence can be hosted by CopilotKit or self-hosted for enterprise deployments. An open-source example application called OpenTag demonstrates a complete Channels implementation using a Python LangGraph agent connected over AG-UI with native Slack and Microsoft Teams experiences.
Key facts
- Channels SDK connects AI agents from multiple frameworks (LangGraph, CrewAI, Mastra, Pydantic AI, Google ADK) to Slack, Teams, and Discord
- Agents can understand conversations, stream responses, call tools, work with files, and render interactive UI with human approval gates
- Messages render once as platform-native UI: Slack Block Kit, Teams Adaptive Cards, or platform-specific interfaces
- Developer infrastructure runs the agent logic and credentials; CopilotKit Intelligence manages platform connections and credentialed delivery
- Setup via CLI command or manual steps; requires Node.js 22+, long-running Node process or container
- MIT licensed open source with self-hosting option for enterprise deployments