GitHub has unveiled a desktop appl designed to act as a centralised control centre for artificial intelligence (AI) agents. Announced at the Microsoft Build 2026 conference, the new GitHub Copilot app is a shift from passive coding assistants to independent AI agents capable of managing complex development workflows in parallel.
The launch comes at a time when there is high demand in automated coding activity. According to internal GitHub data, code commits on the platform have nearly doubled year-over-year, now crossing 1.4 billion per month, while automated GitHub Actions usage has surged past 2 billion minutes per week.
Central control center for developers
While the rise of AI agents has accelerated software development, it has also created a disjointed experience for engineers who must constantly switch contexts, track multiple automated tasks and manually review large amounts of agent-generated code.
The new GitHub Copilot app, currently available in technical preview for Copilot premium tiers, introduces a “My Work” dashboard. This view allows developers to monitor and direct multiple AI agents simultaneously across different repositories. To prevent agents from interfering with one another or messing up local files, every active session automatically runs inside an isolated Git worktree, removing the need for manual branch management.
Visual workspaces and secure sandboxes
Alongside the desktop app, GitHub introduced several core features aimed at improving collaboration between humans and AI. One of those is Canvases – interactive, two-way visual surfaces that display the agent’s current plan, terminal outputs, deployments or browser sessions in real time, allowing engineers to edit, reorder, or redirect the agent's work directly.
To ensure security, agents are given bounded environments to test and run code without impacting actual production systems. Organisations can choose to run these sandboxes locally with restricted file and network access, or in fully isolated, ephemeral cloud-hosted Linux environments.
Then there is Agent Merge, an automation tool that monitors continuous integration (CI) pipelines, tracks required human reviewers, fixes failing checks, and automatically merges code once all pre-set organisational conditions are met.