What is the GitHub Copilot app and why it was created

Over the last three years, GitHub Copilot has become the most widely used AI assistant for developers. According to the Gartner Magic Quadrant 2026, GitHub has been the leader among AI coding agents for the third year in a row. However, with the rise of agentic development, a new problem emerged: developers are losing track. When several agents run in parallel — one fixing a production bug, another implementing a new feature, and a third handling review — context gets fragmented across terminals, editors, and browsers.

The new app solves this chaos. GitHub Copilot app is a desktop client for macOS, Windows, and Linux, available as a technical preview for all paid Copilot plans (Pro, Pro+, Business, and Enterprise). Its core is the "My Work" view — a unified dashboard that shows all active agent sessions, issues, pull requests, and automations across connected repositories.

Isolated environments for each agent

Each agent session runs in its own git worktree — an isolated copy of a branch where the agent works without affecting other running tasks. The app manages all worktrees automatically: creating them, managing them, and cleaning them up after completion. The developer doesn't need to manually switch branches or resolve conflicts between parallel tasks.

For power users, GitHub also introduced Copilot Max — an extended plan with higher limits for intensive agent workflows. GitHub hasn't announced the exact price yet; it will be an add-on on top of existing subscriptions.

Canvas: When chat isn't enough anymore

One of the most interesting new features is Canvases — bidirectional workspaces where people and agents collaborate in real time. While chat is good for giving instructions, Canvas shows the actual work: the task plan, pull request status, terminal session, or deployment dashboard. The agent continuously updates the canvas, and the developer can edit, approve, or redirect the work directly on the canvas.

This is the first step towards what GitHub calls the "agent experience" (AX) — a new type of user interface where the human isn't just a prompt issuer but an active director of the entire development process.

Agent Merge: The end of manually fixing CI failures

The biggest practical impact on developers' daily work will likely come from the Agent Merge feature. It monitors the CI pipeline, detects failures, downloads logs, resolves merge conflicts, and automatically pushes fixes — all without developer intervention. You can set how far Copilot should go: just fix tests, incorporate review feedback, or directly merge when all conditions are met.

For DevOps teams, this means a significant shortening of the cycle from opening a pull request to deployment. Anyone who has ever spent an hour hunting down the cause of a broken CI pipeline knows exactly how much time this saves.

A secure sandbox for agents

So agents can not only propose code but also run tests and iterate, the Copilot app offers two sandboxing modes:

Local sandboxing runs directly on the developer's machine in an isolated environment with restricted access to the file system, network, and system resources. Policies can be centrally configured and enforced — something corporate IT departments will especially appreciate.

Cloud sandboxing provides a fully isolated, ephemeral Linux environment hosted directly by GitHub. The advantage is the ability to continue working from anywhere and on any device — sessions sync across desktop, mobile, and browser.

Code review that handles the agentic load

As the volume of agent-generated code grows, so does the pressure on code review. GitHub is responding with an improved Copilot code review, which now offers:

• Medium tier review — uses a more powerful model for more precise analysis on critical repositories. Administrators can set the level to "low" for less risky code and "medium" for key projects.
• /security-review — specialized security analysis of code
• /rubberduck — a generally available feature that uses multiple model families for independent critique of implementations and finding hidden bugs

Important news for companies using Azure DevOps: Copilot code review is now available natively in that environment as well.

Copilot SDK: One runtime for everything

GitHub also announced the general availability of the Copilot SDK in six languages: Node.js/TypeScript, Python, Go, .NET, Rust, and Java. The SDK exposes the same agent runtime that the Copilot app itself runs on. Companies can thus build their own tools — internal code analyzers, release notes generators, or support agents — on the same foundation, instead of assembling their own infrastructure.

The Copilot CLI also received improvements: a redesigned text user interface, voice input (with on-device speech processing), and the ability to schedule recurring tasks via /every.

Comparison with the competition

GitHub isn't alone in the agentic development space. Anthropic launched a redesigned Claude Code desktop client with parallel agent sessions for Mac and Windows in April 2026. Cursor 3 arrived with its Agents Window, which handles parallel agents across local, worktree, cloud, and SSH environments.

GitHub has one crucial advantage over the competition: integration. Developers already have their code, issues, pull requests, CI pipelines, and branch protection rules on GitHub. The Copilot app doesn't require moving context to a new tool — it starts where the context already exists.

GitHub itself sees over 1.4 billion commits monthly (nearly double year-over-year) and over 2 billion minutes of GitHub Actions consumed weekly. The platform running all of this has a natural head start for agentic workflows.

Pricing and availability in the Czech Republic

GitHub Copilot app is included in existing paid plans — there is no separate charge for it:

• Copilot Pro: $10/month (approx. 220 CZK)
• Copilot Pro+: $19/month (approx. 420 CZK)
• Copilot Business: $19/user/month
• Copilot Enterprise: $39/user/month
• Copilot Max: pricing not yet announced

For Czech developers and companies, the app is fully available — just have an active Copilot subscription and download the client from the official GitHub website. The user interface is in English; Czech is not yet supported, which is standard for developer tools. However, Copilot itself understands Czech prompts — you can assign tasks in Czech and generate code with Czech comments.

What this means for Czech developers and companies

For Czech software companies — from startups to enterprises — the Copilot app represents an interesting evolution. Agent Merge can realistically shorten the time from committing code to deployment. Instead of mechanically fixing CI failures and waiting for reviews, the developer takes on the role of a supervisor who directs agents and checks their output.

At the same time, caution is warranted. Trust in an agent that automatically merges code is something teams need to build gradually. GitHub allows this to be configured in steps — from assistance through semi-automatic mode to fully autonomous merging when defined conditions are met.

For teams that already use GitHub Copilot in VS Code or JetBrains IDEs today, the new app is a logical extension — not a replacement for the editor, but a complement for moments when you need to manage multiple parallel tasks at once.