GitHub today launched a platform for managing artificial intelligence (AI) agent workflows across the entire software development lifecycle (SDLC).

Announced at the GitHub Universe 2025 conference, Agent HQ makes it possible to unify the management of AI agents developed by not just GitHub but also third-party providers, including Anthropic, OpenAI, Google, Cognition and xAI. As part of that effort, GitHub is also making it possible for DevOps teams to build and manage AI agents via Agent HQ that provides access to a mission control tool through which they can centrally manage AI agents and track metrics.

GitHub COO Kyle Daigle told conference attendees Agent HQ will make it simpler to employ AI agents that are capable of not just executing tasks but also reasoning about how best to organize and implement them. The goal is to make it possible to organize a fleet of AI agents in a way that reduces the overall amount of cognitive overhead that developers would otherwise encounter, he added.

For example, a plan mode in GitHub Copilot can now be used to build a step-by-step approach for building an application while at the same time being able to more granularly enforce controls, noted Daigle. Additionally, there is also now an AI agent to review code residing in a GitHub repository. Each repository can now have a tailored set of code reviews that can also be integrated with Code QL, an open-source semantic code analysis engine developed by GitHub.

GitHub is also adding branch controls that provide granular oversight over when to run continuous integration (CI) processes and other checks for agent-created code, identity management tools, one-click merge conflict resolution capabilities and integrations with Slack and Linear platforms used within DevOps workflows

Collectively, these capabilities are critical because as new tool chains are created using AI coding tools and agents that are embedded within the VS Code integrated developer environment (IDE), organizations need to be able to ensure quality is not being compromised, said Daigle. “Agents and developers working side by side is not something that’s coming,” he says. “It’s here.”

Mitch Ashley, vice president and practice lead for software lifecycle engineering at the Futurum Group, said as more application development teams embrace AI we’re witnessing the arrival of a software development future that is seeing a shift from writing code in an editor to directing the agentic workflow.

GitHub Agent Hub, integrated with VS Code, makes it possible for developers to orchestrate multiple specialized agents instead of typing every line of code, he added. Developers instead define intent, guide priorities, and review the results, said Ashley. This represents the natural evolution of the coding environment that is managed by a single, intelligent hub, noted Ashley.

It’s still early days so far as the adoption of AI agents in software engineering workflows is concerned, but given the adoption of previous generations of AI coding tools, adoption is likely to accelerate. According to GitHub, AI is already being pervasively employed across the software development lifecycle, with 80% of new developers on GitHub leveraging Copilot within their first week.

The challenge now is moving beyond simply generating code to increasing the percentage of AI code that can be successfully deployed in a production environment. Much of the code currently generated by AI coding tools, unfortunately, is overly verbose in a way that adversely impacts application performance while simultaneously increasing technical debt. Sometimes referred to as AI slop, much of that code also contains known vulnerabilities that can be easily exploited.

Each organization will, of course, need to weigh the tradeoffs but, hopefully, in the age of agentic AI the overall quality of the code being generated by these tools is about to substantially improve.