AIToday
Large Language ModelsAI Coding AssistantsSemantic Kernel BlogPublished: Aug 5, 2026, 04:00 JST4 min read

GitHub Copilot Agent now stable for .NET and Python

GitHub Copilot Agent now stable for .NET and Python

Key takeaway

  • Microsoft released the GitHub Copilot Agent, now stable for .NET and Python, enabling developers to build production-ready coding agents that combine Copilot's built-in capabilities with Agent Framework's governance and observability.

  • The agent can execute shell commands, read and write files, fetch URLs, and integrate Model Context Protocol (MCP) servers and custom tools — all routed through permission handlers that require explicit approval before sensitive actions run, making it safe for enterprise use.

3 Key Points

  1. What happened

    Microsoft announced the GitHub Copilot Agent is now released and stable for both .NET and Python, letting developers use GitHub Copilot's agentic harness as the execution engine for their agents while keeping Agent Framework's extensibility, observability, and approval workflows.

  2. Why it matters

    Developers building coding agents can now combine Copilot's built-in capabilities — shell execution, file read/write, URL fetching, and MCP tools — with Agent Framework's enterprise governance and middleware, without choosing between the two. This opens new workflows for code review assistants, repository maintenance agents, and developer copilots, all governed by permission handlers so agents only do what you explicitly allow.

  3. What to watch

    Developers building agents will need GitHub Copilot CLI installed and an active GitHub Copilot subscription; .NET requires .NET 8+, Python requires 3.11+. Permission handlers control sensitive actions (shell commands, file writes, URL fetches, MCP calls) before they run.

In Depth

Read the full story

Microsoft today released the GitHub Copilot Agent as stable for both .NET and Python, marking the completion of an integration that gives developers a unified way to build production-ready coding agents. The GitHub Copilot Agent is an Agent Framework agent backed by the GitHub Copilot CLI and SDK, where Copilot owns the agent loop (model calls, tool invocation, planning, and session state) while Agent Framework provides a consistent surface for instructions, tools, streaming, middleware, observability, and human-in-the-loop approval.

Developers can use the agent to run shell commands, read and write files, fetch URLs, and invoke Model Context Protocol (MCP) servers — both local (stdio) and remote (http) — such as a filesystem server or the Microsoft Learn documentation API. Every capability is gated by a permission handler that developers provide; the handler receives each request and returns an approve/deny decision, ensuring the agent only does what is explicitly allowed. For example, developers can prompt for approval on shell commands, allow file reads silently, and always deny URL fetches — or approve all trusted operations and prompt only on custom tools marked with approval_mode="always_require" in Python or wrapped in ApprovalRequiredAIFunction in .NET.

The integration handles session management automatically. Developers can reuse a session to keep context across turns and resume an earlier conversation by its session ID, even from a new agent instance. Custom instruction directories can be pointed to for project-specific or team-shared guidelines. The agent participates in Agent Framework's OpenTelemetry tracing, giving operators the same traces and telemetry as every other agent in their system.

To get started, developers need an authenticated GitHub Copilot CLI and an active GitHub Copilot subscription. Installation is via `dotnet add package Microsoft.Agents.AI.GitHub.Copilot` for .NET or `pip install agent-framework-github-copilot` for Python. The CLI is configured through environment variables (GITHUB_COPILOT_CLI_PATH, GITHUB_COPILOT_MODEL) or directly in code via CopilotClientOptions in .NET or GitHubCopilotOptions in Python. This release opens new workflows for code review assistants, repository maintenance agents, developer copilots, and software engineering workflows, all backed by Copilot's coding capabilities and Agent Framework's extensibility and governance.

Context & Analysis

The GitHub Copilot Agent addresses a gap developers have faced: they want to build agents that reason about code, modify files, execute commands, and work across repositories, but GitHub Copilot already provided a powerful coding harness for these tasks while Agent Framework offered the extensibility, governance, and observability features needed for production use. Rather than force developers to choose between them, Microsoft integrated the two by making the GitHub Copilot agent a native Agent Framework provider. This means developers can now leverage Copilot's CLI-backed capabilities — its built-in planning loop, tool invocation, shell access, and file manipulation — as an execution engine while retaining Agent Framework's middleware, streaming, human-in-the-loop approval, and observability through OpenTelemetry tracing.

The release emphasizes production-readiness through governance controls. Because agents have system-level abilities, they must be safely controlled; the permission handler model ensures every sensitive operation (shell commands, writes, fetches, custom tools requiring approval) routes through a developer-provided gate before execution. The SDK enforces this through pre-tool-use hooks and warns if a custom hook would bypass approval. This design lets organizations run agents in production by approving trusted operations and denying or prompting on others, while maintaining consistent observability across different agent providers in their system.

FAQ

What programming languages does the GitHub Copilot Agent support?
The GitHub Copilot Agent is now released and stable for both .NET (requiring .NET 8+) and Python (requiring 3.11+).
What built-in capabilities does the GitHub Copilot Agent have?
The agent comes with shell execution (run commands and scripts), file operations (read existing files and write new ones), URL fetching (pull in and process web content), and integration with Model Context Protocol (MCP) servers.
How does approval work for agent actions?
Every sensitive action — shell commands, file writes, URL fetches, MCP calls, and approval-required function tools — flows through a permission handler you provide, which can approve, deny, or prompt per request; by default nothing runs without oversight.
Semantic Kernel BlogRead Original Article

Get the latest Large Language Models news every morning

AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.

Free · takes 30 seconds · unsubscribe anytime

Ask AI

Ask AI anything about this article. Q&As are published on this page for other readers too.

Related Articles

Next articleCloudflare launches AI shopping ID and wallet service

The AI news that matters, in one minute each morning.

Sign up free