Microsoft is readying a pivotal update to Microsoft 365 Copilot Chat that will let organizations turn their centrally governed, specialized agents into reusable building blocks for multi-agent workflows. According to a newly published entry on the Microsoft 365 roadmap (ID 567670), agents published to Agent 365 will be connectable inside Copilot Chat and declarative agents starting in August 2026, provided they support the Agent-to-Agent (A2A) interoperability protocol.
What’s Changing in Agent 365 and Copilot Chat
The roadmap entry, marked “in development” with a general availability target of August 2026, details a feature that bridges Agent 365—Microsoft’s catalog for publishing and administering agents through the Microsoft 365 admin center—with the builder tools in Copilot Studio and Agent Builder. Once enabled, an agent that an organization has already published to Agent 365 can be selected as a “connected agent” when someone creates a new agent in either tool.
This isn’t a simple API call. Microsoft is requiring that agents implement the A2A protocol, an emerging standard for agent interoperability that allows one AI agent to delegate tasks and share context with another. A connected agent can hand off a job, receive the results, and incorporate them into a larger workflow. In practice, that means a developer building a customer-service copilot could plug in a dedicated sentiment-analysis agent that’s already been approved, governed, and tuned centrally—without rewriting any of its logic.
The feature will work on any platform where Copilot Chat runs: Android, iOS, desktop, and the web, and it will be available in Microsoft’s standard worldwide multi-tenant cloud.
What the A2A Requirement Means for Your Agents
Publishing an agent to Agent 365 will not automatically make it show up as a connected option. The A2A protocol is the gatekeeper. Microsoft’s Copilot Studio documentation already describes how to build agents that can act as A2A-connected external services, but this roadmap entry extends that capability into the Copilot Chat and declarative-agent landscape.
For developers, the implication is clear: to qualify, your agent must expose A2A endpoints that allow it to receive tasks, process them, and return structured responses. The A2A protocol, originally proposed by Google and later formally adopted by Microsoft, defines a standard envelope for inter-agent communication. It specifies how agents advertise their capabilities, accept delegation requests (like a “SendTask” operation), and report status or artifacts. This is more than a conventional REST call—it is a full task lifecycle handoff with context and security tokens.
Administrators, meanwhile, retain control through the admin center, where they can manage publication, security policies, and data boundaries just as they do today. This dual obligation—developer instrumentation plus admin governance—is what will make centrally published agents trustworthy enough to be reused across many solutions.
Why This Matters for Different Microsoft 365 Audiences
For everyday users, there’s no immediate change in your daily Copilot Chat experience—the feature won’t land until August 2026 at the earliest. When it does, though, you may notice Copilot Chat becoming more capable at complex, multi-step tasks because it can quietly offload sub-tasks to specialized agents that already exist elsewhere in your organization. Instead of a single monolithic response, the assistant might orchestrate several behind-the-scenes experts to assemble a more complete answer.
For power users and makers who build agents in Copilot Studio or Agent Builder, you’ll gain a library of pre-approved, governed agent services to snap into your creations. This means you won’t have to reinvent common capabilities like translation, compliance checks, or data analysis. You can focus on stitching them together into a tailor-made workflow, confident that each building block meets your company’s standards for security and data handling.
For IT administrators and governance teams, reuse is the headline benefit. Rather than auditing dozens of independently built agents that each contain a copy of the same sensitive logic, you can manage one canonical agent and let it be consumed across multiple projects. This reduces risk, simplifies compliance, and makes it easier to enforce data loss prevention policies. However, you’ll need to start identifying which existing Agent 365 assets could serve as shared services and verify that their developers are targeting A2A compliance. The August 2026 timeline is a target, not a hard promise, but preparing early avoids a last-minute scramble.
For developers who build agents for their organization or for the Microsoft commercial marketplace, A2A readiness will become table stakes. Look at the protocol’s specification (published by Microsoft and Google) and understand how your agent can advertise its capabilities, accept delegation requests, and return results in a standardized format. Copilot Studio’s existing multi-agent documentation offers a preview of the patterns you’ll need to follow.
How We Got Here: The Multi-Agent Evolution Inside Microsoft 365
Microsoft’s Copilot stack has been steadily inching toward a multi-agent architecture. Initially, Copilot Chat and declarative agents relied on simple extensions and skills—single-purpose plugins that fetched data or performed discrete actions. Copilot Studio later introduced the ability to chain actions and reference external APIs, but each agent remained a self-contained island.
In 2025, Microsoft began publicly embracing the A2A protocol, publishing a draft specification and building support into Copilot Studio for “A2A-connected external agents.” This allowed a Copilot Studio agent to call out to another agent that spoke A2A. The missing piece was bringing that same interoperability to the broader Copilot Chat experience and to the Agent 365 catalog—precisely what the new roadmap entry addresses.
The timing aligns with a broader industry push toward agentic AI, where networks of specialized AI agents collaborate on complex tasks. Google’s own Agent2Agent efforts and emerging standards like MCP (Model Context Protocol) underscore that no single model can do everything; instead, the future belongs to orchestrated teams of agents. Microsoft’s move to make Agent 365 the central clearinghouse for these agents inside its ecosystem positions the admin center as a governance hub for AI at scale.
What You Should Do Now
Even though August 2026 feels distant, there are concrete steps that admins and developers can take today:
- Audit your Agent 365 catalog. Identify which published agents perform functions that could benefit multiple teams—think specialized analysis, compliance checks, or data connectors.
- Check A2A readiness. If you have custom agents built with Copilot Studio or custom code, review the A2A protocol documentation. Determine how much effort it will take to add the required endpoints and message formats. Examine your agent’s current integration pattern; if it relies on direct API calls, you’ll need to layer an A2A wrapper that implements operations like SendTask and GetTaskStatus.
- Start a governance conversation. Decide who should own shared agents, what approval processes are needed before an agent becomes a “connected” service, and how you’ll monitor usage and data flow.
- Test in Copilot Studio’s developer environment. The existing A2A support for external agents in Copilot Studio can serve as a sandbox to prototype how your Agent 365 agent will behave when connected.
- Educate your builder community. Power users and citizen developers who use Agent Builder should understand that reusable agents are coming, and they should plan their designs to leverage them rather than duplicating effort.
No action is required from end users at this stage; the change will happen transparently behind the scenes as their IT departments adopt the new capability.
Outlook: From Specialists to Orchestra Conductors
August 2026 is more than a year away, and Microsoft’s roadmap dates are notoriously fluid. Still, the inclusion of this item in the public roadmap signals a firm direction. The combination of Agent 365, A2A, and Copilot Chat points toward a future where users don’t just chat with a single AI—they interact with a symphony of agents that collaborate under the hood.
For organizations that invest early in the A2A standard, the payoff could be substantial: lower development costs, stronger security posture, and Copilot experiences that finally deliver on the promise of understanding and automating complex, multi-system business processes. Keep an eye on subsequent roadmap updates and any preview programs that might emerge in the months ahead.