GitHub Copilot Agent Gets Smarter

GitHub has rolled out significant enhancements to its GitHub Copilot coding agent, transforming it from a background assistant into a more autonomous and secure development partner. These updates, detailed on the GitHub Blog, include a model picker, self-review capabilities, built-in security scanning, custom agents, and improved CLI handoff.

StartupHub data

Companies working on this

Profiles of the companies named in this story, with funding and a one-liner from our database.

GitHub Advanced Security

GitHub's suite of security tools for code scanning, secret detection, and dependency review.

Designed to handle delegated tasks like bug fixes, test generation, and code cleanup, the Copilot coding agent now delivers pull requests that are closer to production-ready. These recent advancements aim to streamline developer workflows and reduce manual oversight.

Smarter Task Execution and Review

Developers can now select specific AI models for different tasks via the Agents panel. This allows for optimization, using faster models for routine work such as unit tests, or more robust models for complex refactors. This granular control, currently available for Copilot Pro and Pro+ users, ensures efficient resource allocation.

A critical upgrade introduces self-review, where the agent uses Copilot code review to iterate on its own changes before opening a pull request. This process, observed to catch and rectify issues like overly complex string concatenations, significantly reduces the cleanup burden on human reviewers, delivering higher-quality code from the outset.

Enhanced Security and Customization

Security is now integrated directly into the agent's workflow. The Copilot coding agent automatically runs code scanning, secret scanning, and dependency vulnerability checks, flagging potential issues like committed API keys or known CVEs before a pull request is even created. This proactive approach, including GitHub Advanced Security features at no additional cost for agent-generated code, shifts security left in the development lifecycle.

For teams requiring consistent adherence to specific processes, custom agents are now available. By defining an agent file under .github/agents/, organizations can codify unique workflows, such as a performance optimizer that benchmarks, applies changes, and measures improvements before submission. This standardization can be shared across an enterprise, ensuring consistent quality and methodology.

Seamless Workflow Integration

Context switching between cloud and local environments is also improved with CLI handoff. Developers can pull a cloud session into their terminal, retaining the branch, logs, and full context, or push work back to the cloud from the CLI. This ensures continuous flow, preventing loss of progress or conversational context.

These updates collectively position the GitHub Copilot coding agent as a more sophisticated and integrated tool, actively contributing to code quality, security, and developer productivity. Future developments are expected to include private mode, advanced planning capabilities, and expanded non-pull request tasks like issue summarization.