Skip to Main Content

VISUAL STUDIO CODE

Visual Studio Code 1.129 adds dedicated agent host

Microsoft updates Visual Studio Code with a specialized agent host process and an integrated editor panel to streamline AI assisted development workflows.

Read time
4 min read
Word count
851 words
Date
Jul 21, 2026
Summarize with AI

Microsoft has launched Visual Studio Code version 1.129 to improve how developers interact with AI agents. This update features a dedicated agent host process that allows session sharing across multiple windows. Users can now run terminal commands directly from chat using a specific prefix. The release also includes an integrated editor panel within the Agents window for better code reviews. Additionally a preview of a modernized user interface is available as an experimental setting for those looking to test the latest workbench design.

Visual Studio Code 1.129 adds dedicated agent host. Visualization by Stable Diffusion
Visualization by Stable Diffusion
🌟 Non-members read here

Microsoft released Visual Studio Code 1.129 on July 15 to enhance how programmers manage artificial intelligence tools within their workspace. The update introduces a specialized agent host process and a redesigned panel for the Agents window. These changes aim to make AI interactions more stable and deeply integrated into the standard coding environment.

Evolution of Agent Session Management

The core of this update involves a significant rearchitecting of how agent sessions function. Microsoft moved these sessions to a dedicated process known as the agent host. This host handles various agent harnesses including Codex, Copilot, and Claude. By utilizing the Agent Host Protocol, the software ensures that these intelligence tools operate independently from the primary editor interface.

Running a session in its own process provides several practical advantages for developers. Because the session is decoupled from a single window, a programmer can connect to and view the same session from multiple windows simultaneously. This flexibility allows for complex workflows where different parts of an AI conversation are visible in different workspace areas.

The agent host utilizes the Copilot SDK for its primary internal agent. This alignment ensures that the tool behaves consistently with other ecosystem products like the standalone GitHub Copilot application and the command line interface version. Users can activate this new architecture by toggling a specific setting in their preferences. This rollout covers both the main editor window and the specialized Agents window.

Standardizing the behavior of these agents across different platforms simplifies the learning curve for teams. When the agent host functions the same way in the editor as it does in a terminal, developers encounter fewer surprises. This consistency remains a priority as the software evolves to handle more complex automated tasks and multi-file code generations.

Enhancements to the User Interface and Workflow

The Agents window received a substantial visual overhaul in this release. Previously, reviewing the changes suggested by an agent required moving between different panels. The new design introduces a docked editor panel that merges the detail area with the primary editor interface. This unified view uses a shared tab bar to streamline the review process.

Reviewing an agent’s work now feels identical to standard file editing. This change reduces the cognitive load associated with switching contexts. Developers can see the conversation with the agent right next to the code changes it generates. This proximity makes it easier to verify the accuracy of AI-driven suggestions before committing them to the project.

Workflow efficiency also improves with the addition of a new chat prefix. By starting a chat message with an exclamation point, users can execute terminal commands directly from the prompt. This feature functions within agent host sessions in both the editor and the Agents window. It bridges the gap between natural language requests and command line execution.

Another significant addition for enterprise users is the support for custom models. Developers can now use their own keys to connect different models within the Agents window. This flexibility allows teams to leverage specific AI providers while still using the integrated tools provided by the editor. The update also includes an experimental preview of a modern user interface for the entire workbench.

Stability and Cross-Platform Availability

The software remains available for major operating systems including Windows, macOS, and Linux. This release follows quickly after the previous version to ensure that new features reach the community without delay. Microsoft also issued a minor subsequent update to address specific technical issues reported by early adopters.

The 1.129.1 patch specifically fixed a bug where remote agent hosts failed to start correctly. Addressing these connectivity issues is vital for developers who work in containerized environments or on remote servers. High reliability in the agent host process is necessary for the tool to remain a trusted part of a daily programming routine.

The experimental modern UI setting gives users a glimpse into the future direction of the platform. While it remains optional, it demonstrates a move toward a more contemporary aesthetic for the workbench. Developers who enjoy testing the latest design trends can enable this feature through the experimental settings menu to provide feedback.

By focusing on the underlying architecture and the user experience of AI tools, this version sets the stage for more advanced automation. The dedicated host ensures that as agents become more powerful, they do not compromise the performance of the editor itself. Microsoft continues to refine these systems to support the growing reliance on AI assisted coding.

The integration of terminal commands into the chat interface reflects a broader trend toward conversational development. Instead of manually typing long paths or complex flags, programmers can ask the agent to perform the task. This reduces the need to leave the editor environment, keeping the focus on the code and the logic of the application.

With the ability to bring external models into the workspace, the editor becomes a more versatile hub for various AI services. This neutrality allows developers to choose the best tool for a specific task without losing the benefits of the integrated environment. The 1.129 update represents a clear step toward a more unified and capable development platform.

References