Sign in
← AI PulseAug 26, 2026

Policy · news · Single-source brief

Microsoft Agent Framework for Python Introduces Channels for Agent and Workflow Connectivity

Microsoft has introduced channels within its Agent Framework for Python, allowing agents and workflows to connect with various interfaces and systems, including OpenAI Responses clients, Telegram, A2A, and MCP clients.

By Illumora Editorial

Source · Aug 26, 2026, 2:47 PM · On Illumora · Aug 26, 2026, 2:52 PM

Media from the primary source — shown here so you can stay on Illumora.

Rewritten from one allowlisted primary — not independent enterprise reporting. Lanes →

Brief drafted by Illumora’s editorial model from the linked primary source. Ops desk reviews flagged pieces. How we write →

Read the source →Microsoft Semantic Kernel Blog — Microsoft Agent Framework for Python: Connect with Different Channels
Save

Microsoft's Agent Framework for Python now includes new channel packages designed to facilitate agent and workflow interaction across diverse platforms. This update enables builders to manage how their agents are exposed and behave on different channels, without requiring the agent or workflow to be placed inside a new application runtime or maintained with multiple SDKs.

The new packages are structured around a shared hosting core and specific channel integrations. Each channel package focuses on its protocol boundary, while the application retains control over routing, authentication, authorization, storage, background processing, and deployment.

Key Points

  • The agent-framework-hosting package provides a common foundation for channels, utilizing AgentState for agents and WorkflowState for workflows.
  • AgentState maps application-selected session IDs to AgentSession values, with the application responsible for defining how a caller or user becomes a session ID.
  • This session ID mapping allows a single conversation to continue across multiple channels, such as OpenAI Responses and Telegram, if they resolve to the same canonical session ID.
  • Responses helpers convert incoming requests into Agent Framework run values and transform results back into the Responses format.
  • Telegram helpers translate native updates into Agent Framework input and convert streaming runs into Telegram operations, with the application executing these operations.
  • For A2A, AgentA2AAdapter and WorkflowA2AAdapter generate native agent cards and align input/output modes with conversion helpers.
  • For MCP, AgentMCPTool and WorkflowMCPTool derive native tools from Agent Framework targets, supporting integration with servers like FastMCP.

Context

According to the Microsoft Semantic Kernel Blog, the utility of an agent or workflow depends on its accessibility through existing interfaces and channels. The introduction of channels in the Microsoft Agent Framework for Python addresses this need by providing builders with control over channel exposure and agent behavior.

Why It Matters

This update offers builders greater flexibility in deploying and managing agents and workflows. By separating the agent definition from its communication channels, it simplifies the process of integrating agents into diverse application environments and user touchpoints, potentially reducing development overhead and increasing reach.

What To Do

  • Review the Microsoft Agent Framework for Python documentation for details on the new channel packages.
  • Examine the provided samples for Responses, Telegram, A2A, and MCP to understand practical implementations.
  • Note how AgentState and WorkflowState are used to manage agent and workflow targets and sessions across different channels.
  • Consider how to implement application-owned identity linking, authorization, and concurrency controls for shared sessions.