Weekly GitHub Copilot Roundup: Agents, Models, and Governance
This week's updates bring new custom agents, enhanced models, deeper IDE integration, and improved governance tools for developers and teams.
GitHub Copilot Custom Agents
Custom agents are now available in Copilot, extending beyond standard code completion to streamline DevOps, security, and automation workflows. Teams can define agents in markdown and manage them inside repositories. Integration examples include PagerDuty, JFrog, and Neon. These agents, which run in the terminal, VS Code, and on GitHub.com, provide automation for specific domains and support organization-wide policies or coding standards. Tutorials such as Rubber Duck Thursdays show how to build and set up agents tailored to the needs of your team. Strong vendor integrations and accessible setup options enable flexible AI-driven automation for software pipelines.
- Introducing Custom Agents in GitHub Copilot for Developer Workflows
- Rubber Duck Thursdays: Building with Copilot Custom Agents
GitHub Copilot Spaces
GitHub Copilot Spaces now support sharing via view-only links, making it easier to share documentation, reusable code, and learning materials for open-source projects. These Spaces include role-specific controls and only host user-generated content, focusing on a balance between access and security. Another feature supports adding files directly from GitHub’s code viewer, streamlining workspace creation for AI-powered changes. Updated documentation explains how Copilot Spaces can assist with debugging, planning, and onboarding while maintaining privacy and focusing on team efficiency. The features support collaborative, context-rich work for teams.
- Accelerate Debugging with GitHub Copilot Spaces and Copilot Coding Agent
- Major Updates to Copilot Spaces: Public Spaces and Code View Integration
Copilot Model and Chat Enhancements
Recent Copilot updates provide public preview access for OpenAI’s GPT-5.1-Codex-Max and add support for Claude Opus 4.5 in Copilot Chat, delivering expanded model choices. Copilot Pro, Business, and Enterprise users have flexible options for code generation and model selection. Copilot Chat in Visual Studio now includes web URL context, letting users reference and query current online content—helpful for questions beyond the model’s existing data. These features equip teams with stronger models, targeted responses, and more manageable AI interactions.
- OpenAI’s GPT-5.1-Codex-Max Public Preview Release for GitHub Copilot
- Claude Opus 4.5 Preview Available in GitHub Copilot Chat and IDEs
- Unlocking Developer Productivity with Copilot Chat’s New URL Context
GitHub Copilot in Visual Studio 2026
The Visual Studio 2026 update boosts Copilot’s integration by adding a GitHub Cloud Agent and expanded contextual actions. Developers can now handle documentation, refactoring, and batch editing directly through Copilot’s interface. New UI features include one-click actions, smarter code search suggestions (“Did You Mean”), and improved code refactoring and hierarchy visualization for C++ projects. C++ preview enrollment is open to more users, extending Copilot’s toolkit for Visual Studio.
- Visual Studio 2026 Released: GitHub Cloud Agent Preview and Copilot Features
- GitHub Copilot and Visual Studio 2026: November Update Highlights
- Enhancing C++ Development in Visual Studio 2026 with GitHub Copilot
Copilot CLI and MCP Enhancements
Building on recent work around registry and deployment, new tutorials walk through setting up a private registry on Azure API Center, so only trusted models are accessible in Copilot and VS Code. There are demonstrations for the kit-dev MCP Server CLI, including code symbol extraction, abstract syntax tree searching, and inline documentation. The guides help teams securely automate Copilot and MCP tasks using compliant workflows.
- Locking Down MCP: Create a Private Registry on Azure API Center for GitHub Copilot and VS Code
- Supercharging GitHub Copilot CLI with MCP Server
Copilot Agent Automation, Orchestration, and Evaluation
Step-by-step guides continue from last week’s agent orchestration materials, showing how to use Mission Control for Copilot agent assignment, prompt creation, and parallel execution. The ongoing AI Toolkit + Copilot Pet Planner series now covers agent setup, code output generation, iterative tracing, and results evaluation. Tutorials focus on reviewing trace data, comparing agents side by side, and scoring output, making agent development easier to manage.
- How to Orchestrate Multiple GitHub Copilot Agents Using Mission Control
- Setting Up AI Toolkit and GitHub Copilot for Microsoft Foundry Projects
- Generating Agent Code Using AI Toolkit and GitHub Copilot
- Creating an Agent with AI Toolkit and GitHub Copilot: Pet Planner Workshop Part 3
- Adding Tracing to an Agent with AI Toolkit and GitHub Copilot
- Evaluating AI Agent Output with GitHub Copilot and AI Toolkit (Pet Planner Workshop, Part 6)
- AI Toolkit and GitHub Copilot: Model Recommendations Workshop
- Evaluating AI Models for Coding with GitHub Models
Issue Assignment and Project Management Integrations
Now, issues can be assigned directly to Copilot using GraphQL/REST APIs, streamlining automation for code review, triage, and routing CI/CD workflows. Teams can set up custom agent directions and use Copilot with Linear’s issue tracker for automatic code or pull request generation, expanding integrations with other tools.
- Assign Issues to GitHub Copilot Using the API
- Assigning Linear Issues to GitHub Copilot Coding Agent
Administration, Auditing, and Code Generation Metrics
New governance features allow organizations to see more code generation metrics with Copilot Insights Dashboard, breaking down activity by model, user, trigger, and language. Metrics can now be exported, and the Control Panel now provides a unified location for managing agent access, permissions, and audit logs. Better audit trails support secure deployments and help organizations meet compliance requirements.
- Track Copilot Code Generation Metrics in GitHub Insights Dashboard
- Managing and Auditing GitHub Copilot Agents: Insights and Governance Tools
Advanced Copilot Use Cases: Code Review, Performance Profiling, HPC Automation
Pull request review integration now includes automated and custom review features, with CodeQL static analysis. Visual Studio 2026’s Profiler Agent enables natural-language performance analysis using BenchmarkDotNet for .NET projects. For high-performance computing, Copilot helps automate Slurm jobs via GPT-5-based models, reducing manual scripting in scientific workflows.
- Accelerating Pull Request Reviews with GitHub Copilot Code Review
- Optimizing .NET Performance with Copilot Profiler Agent in Visual Studio 2026
- Automating HPC Workflows with Copilot Agents
Copilot Studio Intelligent Agent Development
A Microsoft Ignite session highlighted ways to develop advanced Copilot Studio agents using Microsoft Graph, Azure AI Search, and Active Directory. Teams can use connectors and business logic to filter documents and analyze information, supporting enterprise automation and aligning with Microsoft security standards.