GitHub announced the new Copilot application during Microsoft Build 2026. The tool offers a native desktop experience aimed at developing with AI agents. Desenvolvedores can now track parallel sessions, inspect changes, and manage workflows all in one place. The app is now available in technical preview for users of the Copilot Pro, Pro+, Business and Enterprise plans.
The launch responds to a significant growth in the use of agents. Commits on GitHub has almost doubled year over year and exceeds 1.4 billion per month. GitHub’s Minutos Actions exceeds 2 billion per week. Traditional Ferramentas did not keep up with the multiplicity of agents running at the same time.
Aplicativo brings together agent sessions in central view
The Copilot GitHub app creates a “My Trabalho” view that captures everything. A developer can open the day and see three tasks in progress. An agent investigates a bug in production. Outro implements pending task. A third party reviews pull request feedback.
Cada session runs in its own isolated Git work tree. Isso avoids conflicts between agents. The system manages branches, cleanups and contexts automatically. The user inspects, redirects, tests, or merges changes directly in the interface.
- Sessões active in connected repositories
- Problemas and pull requests in progress
- Automatizações running in the background
- Acesso quick to review plans and systems
Bidirectional Telas allows visual inspection of work
The app introduces screens as work surfaces shared between humans and agents. A screen can show task plan, pull request, endpoint, deployment, or flow dashboard. Agentes update content in real time. Desenvolvedores edit, reorder or approve what appears.
Essa approach solves the problem of long chats full of decisions. The work stops being just text and becomes something visible and manipulable. The chat continues to be used for initial instructions and clarifications. Screens transform intention into concrete artifact.
On-premises and cloud Sandboxes ensure secure execution
Agentes need to execute code, test and iterate. Copilot offers sandboxes for this. The local option runs on the user’s machine with configurable file and network restrictions. Corporate Políticas can be applied centrally.
In the cloud, each sandbox uses an ephemeral Linux environment hosted by GitHub. Organizações define their own rules. Sessões can be resumed on any device. The model prioritizes security and avoids impact on production environments.
Agent Merge automates pull request advancement
Depois from code generation, Agent Merge drives the pull request. Ele monitors continuous integration, tracks reviewers, and resolves failed checks. The user defines limits: how far the agent should go in fixing problems or automatic merging.
Essa functionality reduces the review bottleneck when contribution volume grows. Copilot Code Review complements the process by adapting to the team’s internal standards.
Copilot Max Serves High Volume Users
Advanced Usuários can opt for Copilot Max as an upgrade. The version supports greater intensity of use without interrupting the main flow. Ela maintains compatibility with existing plans and focuses on performance for large teams.
The application integrates with GitHub’s larger strategy of scaling systems to the “agetic computing frontier.” The company reinforces infrastructure at all layers to support increased activity.
What changes for development teams
Equipes gains centralized control over multiple agents. Field Engenheiros can orchestrate initiatives without switching between disjointed windows. The focus returns to human judgment at critical points as agents handle repetitive tasks.
The app does not replace existing code editors. Ele acts as a coordination layer over the GitHub ecosystem. Integrações with repositories, issues and Actions remain native.
Desenvolvedores report reduced time spent on context management. Agent-generated Alterações arrives with a clear record of what has been tried, validated and where human intervention is required.