Google, IBM & Meta Certificates – 40% Off
One Coursera Plus subscription covers most Professional Certificates on Coursera.
Unlock All Certificates
- Explore Microsoft Copilot Studio, compare its three agent harnesses, and select an appropriate harness for a business scenario.
After completing this module, you can:
- Explain the role of a harness in Microsoft Copilot Studio.
- Compare the GitHub Copilot, standard, and Copilot Chat harnesses.
- Identify the components and lifecycle of an agent powered by the GitHub Copilot harness.
- Select a harness based on business and technical requirements.
- Create an agent powered by the GitHub Copilot harness, define its behavior with instructions, add knowledge, and refine it through testing.
After completing this module, you can:
- Create an agent powered by the GitHub Copilot harness.
- Define an agent's purpose, scope, behavior, tone, boundaries, and escalation approach.
- Add file and website knowledge sources.
- Configure agent details, a greeting, and suggested prompts.
- Test and refine an agent in Preview.
- Ground agents in approved information, connect them to external services, and add reusable skills in Microsoft Copilot Studio.
After completing this module, you can:
- Distinguish knowledge, tools, skills, and connected agents, and choose the appropriate capability type.
- Add and validate knowledge sources, grounding, and citations.
- Configure an existing connector tool and its inputs.
- Connect an agent to an existing Model Context Protocol (MCP) server.
- Add a supplied skill and refine its activation description.
- Test connector, MCP, and workflow tools and connected-agent delegation, and improve an agent based on the results.
- Build and connect standalone workflows to agents in Microsoft Copilot Studio. Design input/output contracts, configure connector actions, require end-user confirmation, and test success, missing-information, cancellation, and failure paths.
After completing this module, you can:
- Decide when to use agent reasoning and when to use a deterministic workflow.
- Design workflow trigger inputs and response outputs that support reliable agent-to-workflow communication.
- Create a standalone workflow in the Copilot Studio Workflows experience.
- Configure a connector action using dynamic content from trigger inputs.
- Add a workflow to an agent as a tool and require end-user confirmation before it runs.
- Test success, missing-information, cancellation, and basic failure paths, and use workflow activity to diagnose problems.
- Evaluate agent quality with repeatable test sets, use monitoring data to identify runtime issues, publish agents to Teams and Microsoft Copilot, and apply governance, lifecycle, and Copilot Credit considerations.
After completing this module, you can:
- Distinguish informal testing on the Preview tab from repeatable evaluation.
- Create representative test cases with expected outcomes and review evaluation results.
- Evaluate whether a main agent delegates requests to the correct connected agent.
- Use the Evaluate and Monitor tabs to identify quality, grounding, capability-selection, delegation, and runtime issues.
- Publish an agent to Teams and Microsoft Copilot and validate its authentication, permissions, knowledge, tools, workflows, and connected agents.
- Explain governance and application lifecycle management considerations for agents.
- Identify authoring and runtime activities that consume Copilot Credits and where administrators monitor capacity.
