Copilotアプリが全プラン開放 — BYOKとJetBrainsのCodex連携GitHub opened its standalone Copilot app to all plans—including Copilot Free…

GitHub has extended access to its standalone Copilot application, making the desktop client available across every Copilot plan—including the no-cost Copilot Free tier and GitHub Education—as of July 7, 2026. The change matters because it moves an editor-independent, agent-driven coding environment out from behind a paid subscription and into the hands of free-tier users, while a simultaneous update to JetBrains IDE agent features broadens where these workflows can run.

Standalone desktop clients are not unusual in themselves, but two aspects of this release stand out. The first is that the app now works on Copilot Free, rather than being limited to paid Individual, Business, or Enterprise plans. The second is the addition of BYOK—bring your own key—support, which lets users connect their own model providers without holding a Copilot subscription at all. In practice, that means the application can serve as an execution surface for agentic tasks even for people who supply their own API access to an external model, decoupling the tool from GitHub's own model billing.

The broader significance is architectural. For most of its history, Copilot has lived inside an editor, surfacing completions and chat within Visual Studio Code, JetBrains IDEs, and similar environments. A standalone app repositions the agent as a process that runs independently of any single editor, capable of planning and carrying out multi-step work on a codebase from its own window. Bringing that capability down to the free tier lowers the barrier for developers who want to experiment with agentic workflows without committing to a plan, and it aligns Copilot more closely with a wave of autonomous coding tools that operate outside the traditional editor loop.

BYOK is a useful concept to understand here. In this context it refers to supplying credentials for a model provider you already pay for or host, so the application routes requests to that endpoint instead of to GitHub-managed models. This appeals to users who want control over which model they use, who need to keep inference within a particular vendor relationship, or who are managing costs and data-handling requirements on their own terms. Because the standalone app supports BYOK without a Copilot subscription, the tool can function as a client shell around a user-provided model, which is a notable shift from the more tightly bundled arrangement that has defined Copilot to date.

On the same day, GitHub updated the agent functionality on the JetBrains IDE side, and the release notes point to a Codex integration for that environment. Codex is associated with OpenAI's line of code-focused models and agents, and its appearance alongside the JetBrains update suggests GitHub is continuing to widen the range of model backends and agent behaviors available to developers working in those IDEs. The exact scope of that integration is best confirmed against GitHub's own documentation, but the pairing of a standalone app, BYOK, and a JetBrains-side Codex connection points to a consistent direction: more flexibility in where the agent runs and which models power it.

For context, the free tier itself is a relatively recent addition to the Copilot lineup, having been introduced to give developers a metered, no-cost entry point to the service. Extending the standalone app to that tier is a logical continuation of that strategy, and it places Copilot in more direct comparison with editor-independent and terminal-based agent tools that have gained traction, including command-line coding agents and standalone assistants from other vendors. The presence of BYOK also mirrors an industry pattern in which platforms increasingly let users attach their own model providers rather than locking them to a single hosted option.

The practical takeaway is that developers on any plan can now try an agentic execution environment that does not depend on an editor being open, and that those without a Copilot subscription can still use the app by connecting their own model provider. As always with early releases of this kind, capabilities may differ by plan, region, and platform, and usage limits on the free tier are likely to apply, so checking the current terms before relying on it for production work remains advisable.