News
In the dev team’s new weekly release cadence, Visual Studio Code’s two most recent releases pushed further into agent-centric development, with version 1.115 on April 8 and version 1.116 on April 15 both centered heavily on chat, agents and GitHub Copilot-related workflows. In practical terms, the pair of updates added a new agent companion app, more direct terminal control for agents, better tools for tracing past agent behavior and new controls for tuning reasoning in Copilot CLI.
The bigger story is that Microsoft is extending AI assistance in VS Code beyond inline chat. In these two releases, the editor gained more infrastructure for running, supervising and debugging agent sessions across repos and terminals, while also lowering friction by making GitHub Copilot available as a built-in capability in 1.116 instead of requiring a separate GitHub Copilot Chat extension installation.
Agents App Expands Beyond the Editor
The headline AI addition in the 1.115 release notes was Visual Studio Code Agents, a new preview companion app that ships alongside VS Code Insiders and is described as being built for agent-native development. Microsoft said the app can parallelize tasks across projects by letting developers kick off agent sessions across multiple repositories in parallel, each isolated in its own worktree, while also providing ways to monitor progress, review diffs inline, leave feedback for agents and create pull requests without leaving the app. The release notes also said that custom instructions, prompt files, custom agents, MCP servers, hooks and plugins carry over into the new app, which makes the feature more significant than a standalone experiment because it reuses the same customization layer developers are already building around chat and agents.
Version 1.116 continued that work rather than treating the Agents app as a one-off preview item. Microsoft added several refinements to the app, including reasoning level selection, automatic plan mode handling for CLI sessions involving planning, the Files tab showing by default in Changes, response and rendering improvements, and a rename to Visual Studio Code Agents — Insiders. The release also added a new entry point from the VS Code welcome page, which indicates Microsoft is trying to make the app easier to discover inside the broader product experience.
Agent Debugging and Copilot Controls Get More Practical
The most important AI-focused addition in the 1.116 release may be the new Agent Debug Log workflow. Microsoft said the Agent Debug Log panel shows a chronological event log of agent interactions during a chat session and can now display not only the current session but previous sessions as well, with logs persisted locally on disk. For developers working on prompt files, custom agents, skills or other chat customizations, that turns agent behavior into something that can be inspected after the fact instead of only watched live in a single session.
The same release also added a new way to configure thinking effort in Copilot CLI sessions. According to Microsoft, developers can choose a reasoning model in the language model picker and then select an effort level to balance response quality and latency, with available levels varying by model and no submenu shown for non-reasoning models. That makes the CLI workflow more configurable for teams deciding when they want deeper reasoning versus faster turnaround.
Another notable change in 1.116 was that GitHub Copilot became built in, which Microsoft summarized as letting users start using AI without installing the GitHub Copilot Chat extension. The release notes do not document additional setup details in that highlight line, but the change matters because it reduces one more step between installing VS Code and accessing its AI tooling.
Terminal Tooling Becomes a Core Agent Surface
Across both releases, terminal integration emerged as one of the clearest signs of where VS Code’s AI work is heading. In 1.115, Microsoft added a new send_to_terminal tool so an agent could continue interacting with a terminal after it had moved into the background, addressing a prior limitation where background terminals were effectively read-only apart from output retrieval. The same release also introduced experimental background terminal notifications so the agent can be notified automatically when a background command finishes or requires input.
Version 1.116 broadened that model by extending send_to_terminal and get_terminal_output to foreground terminals too, not just background terminals created by the agent. Microsoft said that means agents can now read from and send input to any terminal visible in the terminal panel, including a running REPL or interactive script. The update also removed an LLM-based prompt-for-input detection step that previously triggered extra model calls on terminal output, replacing that with direct handling through send_to_terminal and user deferral through the question carousel when needed.
Taken together, the two releases show Microsoft concentrating less on isolated AI novelties and more on the mechanics developers need to make agent workflows usable day to day: a dedicated app, persistent debugging, adjustable reasoning in the CLI and deeper control over active terminal sessions. That pattern is more notable than any single UI tweak because it ties chat, agents, CLI and terminal execution into one increasingly connected workflow.
Other Notable Changes in 1.115 and 1.116
The remaining updates were a mix of AI-adjacent workflow improvements, UX changes and a small set of deprecation and fix notes.
About the Author
David Ramel is an editor and writer at Converge 360.
Printable Format
Two Microsoft experts explain how GitHub Copilot is evolving from a coding assistant into a broader platform for building, customizing and testing AI-powered developer workflows.
Microsoft released Visual Studio Code 1.123 on June 3, adding agent-focused features, larger model context support, integrated browser updates and a new delay for some automatic extension updates.
Developer complaints about GitHub Copilot’s new usage-based billing model have centered on unexpectedly rapid AI credit consumption, and neither GitHub nor Microsoft has responded directly to the backlash, though they have previously published guidance to lessen model usage costs.
GitHub’s brand-new Copilot desktop app, in technical preview, handled a small Blazor issue from planning through pull request creation, but the hands-on test also showed why developers still need to verify agent work in the running app before merging.
Subscribe on YouTube
> More Webcasts
Problems? Questions? Feedback? E-mail us.
Leave a Reply