GitHub Copilot evolves into autonomous agent with asynchronous code testing
1 min read
Summary
Microsoft’s GitHub Copilot tool has updated its features to allow users to set up asynchronous code testing.
This means users can assign issues within the tool for it to complete while they work on other tasks.
When invoked, Copilot can navigate and edit code, run commands and open pull requests, embedding into GitHub and cloning repo repositories to decide its workflow and update continuously.
It considers context from previous pull request discussions and follows any custom repo instructions, logging all steps for the human developer to monitor.
This makes the tool more autonomous and competitive in the growing AI coding assistant space, which includes OpenAI’s Codex software engineering agent and Google’s Code Assist.
Support has also been added for Model Context Protocol (MCP) so the agent can retrieve additional data for projects it is reviewing.