Copilot Studio adds triggers, skills and model options

Key insights

• Copilot Studio now groups channels and triggers in one area, according to the demonstration.
• Preview MCP and A2A triggers let other clients or agents interact with an agent.
• The skills marketplace supports sharing skills with teammates and adding them to agents.
• The AI skill generator is presented as suitable for simple skills; document and template support is unclear.
• The demonstration says Cowork is available through the Frontier program, not public preview.

Microsoft Copilot Studio brings channels and triggers together and adds preview options for exposing agents through MCP and agent-to-agent protocols. The changes matter to makers building agents and workflows, and to administrators assessing new agent capabilities before production use. Andrew Hess - MySPQuestions demonstrates several of these updates and describes what he found in his environment.

What's new

The update is a collection of changes rather than one feature. The demonstration covers a skills marketplace, AI-generated skills, model choices, evaluation inside workflows, new trigger types and a Customer Conversation harness. It also shows Managed Runtime Apps and discusses where cost management and Power Platform governance are handled.

One notable interface change is that channels have moved into the Triggers area. The demonstration says recurrence and destinations such as OneDrive for Business, Microsoft Teams and Copilot Chat appear in this combined area. It also calls out preview triggers for MCP and A2A: these let other clients or agents connect with an agent, rather than only having the agent call an MCP server.

Skills are reusable sets of expertise that makers can share with teammates and add to agents. Copilot Studio also offers an AI option to generate a skill. The presenter describes that option as suitable for simple skills, while noting that document or template support was unclear in the version he tested. In workflows, an Evaluate node can be used to measure agent quality and behavior over time.

Who it is for and what you need

These capabilities are relevant to Copilot Studio makers who create agents and workflows, and to teams that want to reuse skills. The material does not specify the licenses, regions or tenant requirements for the Copilot Studio changes, nor does it establish that every feature is available to all customers. Check the linked Microsoft documentation and your own environment before planning a rollout.

Preview status matters. The presenter explicitly describes the MCP and A2A triggers as previews and advises against building them for production. Managed Runtime Apps are also shown in a preview environment. For Cowork, he says it was not in public preview in his demonstration and was available through the Frontier program; that statement should not be treated as a general availability guarantee.

For Managed Runtime Apps, the demonstration says consumption must be enabled in the environment. It identifies Dataverse as the preferred backend for performance, while noting that other backends may be possible. The presenter also says the Microsoft 365 admin center is where Copilot Studio cost management is handled, while Power Platform governance remains in the Power Platform admin center.

How it works

The demonstration focuses on where makers can find and use the features, rather than providing a complete deployment guide. Microsoft documentation linked below covers event triggers, MCP and agent-to-agent channels, skills, model availability and consumption-based billing. Use the following as an orientation, then verify the exact controls in your tenant.

1. In Copilot Studio, open the agent or workflow authoring experience and review the available model choices. The demonstration describes an Auto choice and categories for different task needs; availability may vary.
2. Open the Triggers area to manage channels and trigger options. Where available, inspect the MCP and A2A options, but treat them as preview features rather than production-ready integrations.
3. Use the skills marketplace to share or add an existing skill to an agent. For a new, simple skill, try the AI generation option and review the result before reuse.
4. In a workflow, add or run an Evaluate node to assess an inline agent's quality and behavior as it changes. The material does not provide evaluation criteria or a full configuration walkthrough.
5. For a Managed Runtime App, use the Apps preview entry point shown in the demonstration, enable consumption where required, and iterate on the generated app rather than assuming the first result is finished.

Model selection deserves testing rather than assumption. The presenter recommends trying GPT Chat Latest with evaluations in his Standard harness environment, but that is his advice based on what he saw, not a guarantee that it is the best choice for every agent. The Customer Conversation harness is described as oriented toward voice and chat experiences; the demonstration does not explain its setup in depth.

Limitations and open questions

The clearest limitation is preview status: the MCP and A2A triggers should not be treated as production-ready based on this material. The Cowork access description is tied to the presenter's environment and Frontier access. The demonstration does not establish rollout dates, licensing, regional availability or tenant-wide access for the features discussed.

Some points are explicitly tentative. Backend alternatives to Dataverse are mentioned with uncertainty, and support for documents or templates in generated skills is not confirmed. The presenter also says he has not deeply explored Managed Runtime Apps or the new Customer Conversation harness. These are useful signals to test carefully, not a substitute for a documented support matrix.

For administrators, the practical takeaway is that authoring, governance and cost oversight are discussed in different places. Confirm current portal responsibilities and billing controls against Microsoft documentation before changing administrative processes. The linked documentation provides feature-specific details, but not every announcement in the demonstration has a corresponding source here.

Sources

• Microsoft Copilot Managed Runtime developer documentation
• Add event triggers in Copilot Studio
• MCP and agent-to-agent publication channels
• Manage skills in Copilot Studio
• Create skills in Copilot Studio
• Model availability for agents
• Manage Copilot Studio costs with usage-based billing
• Andrew Hess - MySPQuestions demonstration

Frequently asked questions

Are Copilot Studio MCP and A2A triggers generally available?

The demonstration describes both as preview features and advises against using them for production. Availability details are not stated.

Where did Copilot Studio move channels?

The demonstration says channels are now grouped under Triggers. It does not provide a complete navigation path or confirm the layout across all environments.

Does Copilot Studio AI generate skills from documents?

The presenter tested AI generation for simple skills but was unsure whether that version supported documents or templates. The material does not confirm that capability.

What is needed to use Managed Runtime Apps?

The demonstration says consumption must be enabled in the environment and identifies Dataverse as the preferred backend. Licensing and broader availability are not stated.