Weekly GitHub Copilot Roundup: Agents, MCP, and Model Choice

GitHub Copilot continues to expand its feature set, integrations, and workflows. New agent workflows, model support, and interactive capabilities are providing more ways for developers to automate coding, documentation, and review. The range of updates covers cloud-based agent automation, deeper editor integrations, custom workflow flexibility, enhanced model selection, and new security features—focused on creating a more responsive and productive development environment.

Copilot Coding Agents: Autonomous Workflows and Expanded Integration

Copilot Coding Agents now support more environments, allowing developers to delegate tasks from Visual Studio, VS Code, GitHub, the CLI, and Raycast. Visual Studio 2026 and newer versions support sending Copilot Chat work to agents for asynchronous execution, improving agent-driven pull request creation and notifications that were part of recent IDE and Actions updates. Raycast-based assignment and monitoring add support for cross-platform task delegation, complementing last week’s improvements to GitHub and VS Code workflows. Business and Enterprise users get advanced model picker options for Claude, GPT-5 Codex, and additional model choices. These updates reflect a continued shift toward enhanced, flexible, and secure agent-powered coding.

Model Context Protocol (MCP) and Interactive Agent Experiences

Model Context Protocol (MCP) enhancements offer better agent coordination and extensibility for VS Code and open agent frameworks. Visuals MCP brings interactive UI components into open source, allowing agents to display dashboards, tables, and previews during development tasks. The MCP Registry for Eclipse unifies agent toolchains, while updated preference and task management options let users personalize agent interactions. By supporting markdown-based agents on Azure Functions with Copilot SDK and AGENTS.md integration, cloud distribution and event-driven agent automation become easier.

Expanded AI Model Support: Claude, Gemini, and BYOK

GitHub Copilot now supports more models. Claude Opus 4.6 and Claude Sonnet 4.6 are generally available, while Gemini 3.1 Pro is offered in public preview. The BYOK (Bring Your Own Key) feature in VS Code allows teams to use their own API keys or self-hosted and open source models. These changes reflect ongoing attention to flexible model selection. Some older Anthropic and OpenAI models are being deprecated in line with the focus on performance and updated support.

Custom Agents and Unified Agent Management

With Visual Studio support for both built-in and custom agents, developers now have access to new debugging and modernization workflows, matching recent rollouts of agent skills and .agent.md standards for JetBrains and VS Code. Custom agent roles and project context tools encourage modular and team-based automation. VS Code brings together management for local, cloud, and partner agents, building toward the vision of unified agent handoff and coordination—sometimes called “Agent HQ.”

Agentic Workflows, Continuous AI, and Automation Enhancements

GitHub’s preview of agent-driven repository automation builds on recent markdown/YAML-based automation and Actions improvements. Copilot, OpenAI agents, and secure container features offer new options for fine-grained controls, making complex bot orchestration possible. Agent HQ is moving toward improved security, fleet management, and workflow reproducibility, supporting safer automation for non-critical pull requests while reproducibility practices develop further.

Copilot Integration: VS Code, Zed, and SQL Tools

Copilot now works in the Zed Editor—offering chat, completions, and code suggestions in more places. VS Code adds CLI integration and session tracking to strengthen its feature set. SQL toolkit improvements provide Copilot-powered support inside SSMS, VS Code MSSQL, and Microsoft Fabric Query Editor, simplifying query generation and automation for teams moving from Azure Data Studio.

Developer Workflow Optimization: App Modernization and Prompts

New Copilot migration guides provide support for legacy .NET upgrades—covering containerization, managed identities, and app restructuring. These guides are based on feedback about common .NET and Azure migration scenarios and highlight the practical impact of Copilot automation. The prompts.chat resource offers growing collections of reusable prompt templates to help teams accelerate processes by sharing best practices.

Copilot Usage Metrics and Analytics

Copilot provides organization-wide dashboards and APIs for real-time monitoring of usage and developer productivity. These tools help teams meet compliance policies and give engineering leads better data for analyzing PR cycle times, review speed, and DevOps performance. The new metrics go beyond past status reports or outage analytics, giving actionable insights for daily process improvement.

Octoverse data highlights how Copilot affects the technologies and languages developers prefer, including growth in TypeScript and more adoption of strongly-typed AI integrations. The GitHub Secure Open Source Fund continues to drive improved security, including autofix, scanning, and Copilot-based workflows for open source projects—maintaining last week’s emphasis on audit logs and automated governance.

Other GitHub Copilot News

Visual Studio 2026 introduces built-in support for Mermaid diagramming, integrated with Copilot to help plan architectures and document workflows more easily. These features add to past architectural tooling in the IDE, with stronger Copilot connectivity. Additional updates show steady progress in agent interoperability and model partnerships, moving toward a consistent agentic ecosystem.