In September 2026 VS Code released versions 1.136 through 1.140, adding a set of Copilot‑agent capabilities that tighten the loop from code generation to pull‑request merge. The changes introduce a refreshed Agents window, preview‑only model selection, scheduled automations, Dev Container session launches, and tighter GitHub context integration, all aimed at reducing manual hand‑offs for AI‑assisted development.
Agent Window and Workflow Enhancements
The Agents pane now surfaces controls for automating repetitive steps and for managing session lifecycles. Practitioners can enable agent merge within an active session, allowing the agent to address review feedback, resolve failed checks, handle merge conflicts, and trigger workflow reruns without manual intervention. A new preview feature, HydraFusion, appears in the model picker for eligible users, automatically selecting models and workflows based on the task at hand. Sessions can be marked as done or set to auto‑clean, keeping the list of active agents concise.
Automation Scheduling and Dev Container Integration
Automation templates can now be scheduled on an hourly, daily, or weekly cadence, or invoked on demand. Users start from a built‑in template or supply a custom prompt, enabling repeatable background work such as code linting, dependency updates, or documentation generation. The Agents window also offers a Use Dev Container command, launching an agent inside the project’s container configuration, whether the host is local, remote, SSH, Tunnel, or WSL. This aligns the agent’s toolchain with the exact environment used for builds and tests.
Chat, Context, and Workspace Improvements
Chat interactions have been refined to avoid interrupting the active turn when an agent replies, preserving the flow of human‑to‑human dialogue. GitHub artifacts—issues or pull requests—can be attached to any new session by pasting a URL or using the Add Context UI, bringing relevant metadata directly into the conversation. Sessions now support hierarchical navigation, allowing users to jump between related chats or switch back to the originating conversation with a single link. Workspace handling has been loosened: a generic Copilot chat can start without a folder, then be attached to a local project later, and Codex conversations can be continued across the VS Code and ChatGPT apps without manual copy‑paste.
Related CloudNinjas coverage: AI engineering.
What This Means For Practitioners
These updates shift more of the CI/CD feedback loop into the IDE, meaning AI‑engineers and DevOps teams can prototype, test, and merge code without leaving VS Code. The preview‑only status of HydraFusion and agent merge suggests early adoption should be limited to non‑production workloads while evaluating stability. Scheduling automations inside the IDE may reduce external cron jobs but requires careful monitoring of resource consumption, especially when agents run inside Dev Containers. The ability to attach GitHub context directly to chats simplifies traceability but also introduces a new surface where sensitive issue or PR data could be displayed; teams should consider visibility controls in shared workspaces. Finally, the session‑cleanup and badge notifications help keep the developer experience tidy, but they also rely on background processes that need to be accounted for in operational monitoring.
Practitioners should experiment with the preview features in isolated branches, benchmark any impact on build pipelines, and establish guidelines for when to rely on agent‑driven merges versus manual review. Keeping an eye on future releases for GA status and any security hardening of the new session management APIs will be essential.
