The Microsoft Agent Framework now includes a stable integration with the GitHub Copilot Agent, available for both .NET and Python. This integration allows developers to leverage GitHub Copilot's agentic harness as an execution engine for agents, while utilizing the Agent Framework's extensibility, tooling model, observability, streaming, and human-in-the-loop approval experiences.
Key Points
- The GitHub Copilot Agent is now stable for .NET and Python development.
- It uses GitHub Copilot's agentic harness for execution, including model calls, tool invocation, planning, and session state.
- The Agent Framework provides a consistent surface for instructions, tools, streaming, middleware, observability, and human-in-the-loop approval.
- Copilot's built-in coding capabilities, such as shell execution, file read/write, and URL fetching, are integrated.
- All capabilities are gated by a permission request system, allowing developers to explicitly approve or deny agent actions.
- Developers can extend agents with Model Context Protocol (MCP) servers for additional tools and data, including local (stdio) or remote (http) services.
- Custom functions can be registered as tools, with approval requirements managed through Copilot's native pre-tool-use hook.
Context
According to Microsoft, developers are increasingly seeking to build agents capable of reasoning about code, modifying files, executing commands, interacting with developer tools, and working across entire repositories. While GitHub Copilot offers a coding harness for these scenarios, additional capabilities like observability, middleware, approval workflows, enterprise governance, and broader agent ecosystem integration are often required.
Why It Matters
This integration provides developers with a structured approach to building sophisticated coding agents, offering granular control over agent actions through explicit permission requests and enabling the incorporation of custom tools and external data sources. This can help ensure that agents operate within defined boundaries and meet enterprise requirements for governance and oversight.
What To Do
- Review the Microsoft Agent Framework documentation for .NET or Python to understand the integration specifics.
- Experiment with the provided code examples to implement permission handlers for agent actions.
- Explore configuring Model Context Protocol (MCP) servers to extend agent capabilities with custom tools or data sources.
- Note how to wrap custom functions to require explicit approval before execution, enhancing control over agent behavior.