A new version of GitHub Copilot has just begun rolling out to users, adding a number of new capabilities, such as the addition of agent mode to VS Code.
With agent mode, Copilot can iterate across an entire project, suggesting terminal commands, analyzing run-time errors, and more. “With simple prompts, agent mode takes Copilot beyond answering a question, instead completing all necessary subtasks across automatically identified or generated files to ensure your primary goal is achieved,” the company’s CEO Thomas Dohmke wrote in a blog post.
As an example, GitHub published a demo video in which a developer needs to update a website for runners that allows them to sort races by name, distance, and time. Upon receiving the request, Copilot analyzes the site to determine what needs to be changed, then begins by updating the backend and the UI, then generating unit tests for both, before handing it back over to the user to run those tests.
Agent mode can utilize Claude 3.5 and 3.7 Sonnet, Google Gemini 2.0 Flash, or OpenAI GPT-4o. It was first launched to VS Code Insiders in February, and is now beginning to roll out to all VS Code users, and the rollout should be complete in a few weeks.
The company also announced a public preview for Model Content Protocol (MCP) support in agent mode. MCP is an open standard for connecting data sources to AI tools, and adding it to agent mode will enable Copilot to use different tools to handle different tasks, such as understanding database schema or querying the web.
For example, asking Copilot to “update my GitHub profile to include the title of the PR that was assigned to me yesterday” would prompt the agent to take that request and the list of all available MCP tools and ask an LLM what it should do next based on that information.
“The GitHub local MCP server equips agent mode with compelling capabilities such as searching across repositories and code, managing issues and creating PRs – turning agent mode into a powerful user of the GitHub platform,” Dohmke wrote.
And finally, after adding support for other AI models like Anthropic’s Claude 3.5 Sonnet and Google’s Gemini 1.5 Pro, GitHub is now announcing that requests to models other than Copilot’s base model (GPT-4o) will be considered premium requests and subject to limits, after having been unlimited to this point.
All of GitHub’s paid plans will still offer unlimited requests for agent mode, context-driven chat, and code completions for the base model. Starting in May, Copilot Pro and Copilot Business customers will receive 300 monthly premium requests and Copilot Enterprise customers will receive 1000 monthly premium requests. The company is also adding a new Pro+ plan that gives 1500 monthly premium requests.
If a customer wants to go beyond their monthly allotment, there is a pay-as-you-go structure, starting at $0.04 per additional premium request.
