GitHub Copilot JetBrains Update Adds Codex as an Agent Provider

What has been added

On July 8, GitHub announced a Copilot update for JetBrains IDEs through its official Changelog. The most noticeable change is that Codex has entered public preview as a new agent provider.

Put simply, Copilot inside JetBrains IDEs is expanding beyond recommending code toward choosing an agent provider to handle specific tasks. The update also adds Hooks support in the Customizations editor, richer MCP server management, and support for custom models.

What it means for Codex to join as an agent provider

Codex is a central name in OpenAI’s development-task agent flow. If Codex becomes a provider inside GitHub Copilot, developers can use Copilot’s familiar UI and Codex-style workflows together in the same IDE environment.

The key point here is not that one more model name has been added, but that the agent-selection structure inside the IDE is changing. Going forward, competition among developer tools is likely to move beyond “who does autocomplete better” toward which agents can use which tools and operate under which rules.

Why Hooks and MCP matter together

Hooks are easiest to understand as a mechanism for attaching predefined actions before or after a specific task. For example, a team might check rules before code generation, or automatically connect testing and review steps after a task is completed. MCP server management deals with the pathways that let agents connect to external tools and data through a standard method.

When both are included, developers are not simply telling an agent to “fix the code.” They can more structurally define which tools may be used, which procedures must be followed, and which model should be selected. It is a signal that AI coding is moving from a personal convenience feature toward a team operations feature.

What developers should watch carefully

This update is a public preview. Before applying it broadly to real work projects, teams should organize their policies first. They need to review the scope of code an agent can read, permissions for external MCP server connections, work logs, review responsibility, and usage costs.

In particular, once multiple models and agent providers sit inside the same IDE, what matters is not only output quality but also the ability to track who created which change with which model. The more convenient the automation, the more important rollback and review flows become.

Game Sunakku wrap-up

This news clearly shows Copilot expanding from a simple code recommendation tool into an agent operations environment. Codex as a provider, Hooks, MCP, and custom models may each look like small features in isolation, but together they form an automation structure for work inside the IDE.

In short, the next battleground for developer tools is becoming less about “whether AI gives good answers” and more about “whether AI can use tools to finish work within team rules”. For JetBrains users, it is realistic to view this public preview not only as a feature trial, but also as a chance to check team operating standards.