GitHub Copilot Major Upgrade: Agent Mode and MCP Arrive on VS Code, Taking the Coding Experience to the Next Level!
Attention VS Code users! GitHub Copilot has received a major update with the official launch of Agent Mode and MCP support, integrating top-tier models from Anthropic, Google, and OpenAI. It also introduces the brand-new Pro+ plan and several features moving into General Availability. Discover a smarter, more proactive way to code and make Copilot your ultimate development partner!
Hey, remember the GitHub Copilot we introduced earlier? Fellow developers hustling in VS Code—there’s exciting news for you. Ready? Because Copilot is about to become supercharged, and in a really good way! 👏
Copilot’s new Agent Mode paired with MCP (Model Context Protocol) support is officially rolling out to all Visual Studio Code users. This is no small change—it means your Copilot now gains far greater agency and awareness of the tools and services around it.
On top of that, GitHub is launching the brand-new Copilot Pro+ plan, which includes more premium requests and officially supports top-tier models from Anthropic, Google, and OpenAI. Oh, and there’s more! The Copilot code review agent and next edit suggestions are graduating from preview to general availability!
Sounds cool, right? Let’s break it down and explore what these updates really bring.
Agent Mode Is Here: Not Just Chatting—It Takes Action!
In the past, you may have thought of Copilot as a smart chatbot or a code completion tool. But now, with Agent Mode, Copilot feels less like a co-pilot and more like someone who can take the wheel.
Agent Mode is rolling out gradually to the stable version of VS Code, and it should be available to all users within a few weeks. If you can’t wait, you can enable it manually right now.
Unlike simple chats or cross-file editing, Agent Mode’s core strength is its ability to take action—to turn your ideas into actual code. You just give it a command, and Agent Mode can break down the task, locate or create the necessary files, and step-by-step accomplish your goal. It can even suggest terminal commands, call other tools, and analyze and fix errors on its own. It’s like having a senior engineer by your side—only this one’s AI. 🤖
Since its February debut on VS Code Insiders, developers have used Agent Mode for everything from auto-fixing code to building web apps.
Impressive, right? Behind Agent Mode are powerful models like Claude 3.5 and 3.7 Sonnet, Google Gemini 2.0 Flash, and OpenAI’s GPT-4o. Currently, Claude 3.7 Sonnet powers Agent Mode with a 56.0% success rate on SWE-bench Verified. With ongoing improvements in chain-of-thought reasoning, its capabilities are only going to grow.
Want to try it out? Easy—just update your VS Code to the latest version and enable Agent Mode in the settings!
MCP: A “Smart USB” for Your Copilot Agent
Now let’s talk about MCP (Model Context Protocol). What’s that all about?
Think of it like a “USB port” for AI. Developers use a huge variety of tools daily—from researching, viewing telemetry, managing infrastructure, to writing and debugging code. That entire stack is the engineering stack.
MCP allows Agent Mode to tap into the context and capabilities of these tools. When you issue a command in VS Code using Agent Mode, the model can use MCP to call different tools to get the job done—whether it’s understanding database structure or querying web info.
For example, say you tell Agent Mode: “Update my GitHub profile with the title of the PR I was assigned yesterday.” Agent Mode can query the available MCP tools, ask the LLM what to do next, call GitHub tools to find your assigned PR and extract the title, and then call another tool to update your profile. It works step-by-step until it gets the job done.
This makes Agent Mode more than just a smart assistant—it becomes fully integrated into your dev workflow.
Great news: GitHub already has a thriving and growing MCP ecosystem! Check out this awesome community MCP server list to explore useful MCP servers.
GitHub has also released an open-source GitHub local MCP server, letting you integrate GitHub features (like cross-repo code search, issue management, and PR creation) directly into any MCP-supported LLM tool. That basically turns Agent Mode into a GitHub power user!
More Model Choices, More Flexible “Premium Requests”
Since the last GitHub Universe, GitHub has introduced several new models for chat, cross-file editing, and now Agent Mode. These include Anthropic’s Claude 3.5, 3.7 Sonnet, and 3.7 Sonnet Thinking; Google’s Gemini 2.0 Flash; and OpenAI’s o3-mini.
With these models now generally available (GA), GitHub has introduced a new concept: Premium Requests.
Don’t worry—unlimited access to Agent Mode, context-aware chat, and code completion (currently using OpenAI GPT-4o as the base model) remains unchanged for all paid plans (Pro, Business, Enterprise).
Premium Requests are optional and allow you to use more powerful or specialized models when needed.
- Copilot Pro users: Starting May 5, 2025, get 300 Premium Requests per month.
- Copilot Business users: Starting between May 12–19, 2025, get 300 Premium Requests per month.
- Copilot Enterprise users: Starting between May 12–19, 2025, get 1000 Premium Requests per month.
- Until then, use these premium models unlimited times—so try them out while you can!
There’s also a brand-new Copilot Pro+ plan, tailored for individual developers needing top-tier performance. It offers 1500 Premium Requests per month and access to top models like GPT-4.5, at $39/month.
All Copilot paid users¹ can also choose a pay-as-you-go option to purchase extra Premium Requests. Individuals and orgs can keep using premium models after their monthly quota, set spending caps, and manage costs easily. Admins for Copilot Business and Enterprise plans can manage requests via billing settings in the Copilot admin portal. Additional requests start at $0.04 each.
Different models consume different numbers of Premium Request credits. This gives you flexibility to choose the best model for your task—while still enjoying unlimited access to the base models.
To learn more about how seat-based Premium Requests work, check out the official documentation.
These Great Features Are Now GA Too!
This update isn’t just about Agent Mode and MCP. Two other incredibly useful features are also now generally available:
- Copilot Code Review Agent: Hugely popular during its preview phase—over a million developers used it on GitHub within just a month!
- Next Edit Suggestions: Let’s you
tab
your way to code glory! It predicts your next likely edits, making coding faster and smoother. 🏆
All in all, GitHub Copilot is becoming smarter, more proactive, and even more integrated into your dev workflow. With Agent Mode and MCP, a new era of AI-assisted development begins.
So what are you waiting for? Update your VS Code, enable the new features, and experience the future of development firsthand!
¹ Includes users of Copilot Pro, Copilot Business, and Copilot Enterprise plans.