Visual Studio Code 1.140, published on September 30, 2026, shifts Copilot further toward managed agent execution. Microsoft’s release notes describe a dedicated Copilot harness, experimental multi-folder sessions, remote task delegation, HydraFusion orchestration, and an enterprise control for the default Auto model-selection tier.
For CTOs and engineering leaders, the signal is clear: agent-first development now requires decisions about workload isolation, execution placement, model orchestration, and enterprise policy. These are platform concerns, not editor preferences.
1. Multi-Folder Sessions Isolate Agent Work
What it is
Microsoft says VS Code 1.140 adds experimental support for multi-folder sessions. Different chats can use separate folders or worktrees, isolating associated terminals, tasks, code changes, and pull-request state.
This is not another chat tab. It is a boundary around the working context of an agent session.
Why it matters
Concurrent agent work can create operational ambiguity when sessions share files, terminals, or task state. The multi-folder model gives platform teams a mechanism to test whether separate worktrees improve control over parallel changes.
Why this matters: Your team should evaluate the feature against its existing branch, worktree, code-review, and pull-request practices. Experimental isolation should not be treated as a substitute for those controls.
A useful pilot would compare a shared-folder workflow with separate folders or worktrees, then examine where developers can clearly attribute terminals, tasks, and code changes to the correct chat. The decision is whether this separation reduces coordination overhead without fragmenting review.
2. The Copilot Harness Standardizes Agent Sessions
What it is
According to Microsoft’s VS Code 1.140 release notes, GitHub Copilot agent behavior now runs through a dedicated harness. Microsoft says the change is intended to make agent sessions behave consistently across VS Code and other Copilot tools.
The harness is the architectural story behind the visible workflow changes.
Why it matters
Consistency across tools can make agent workflows easier to operationalize, but vendor-reported consistency still requires local validation. Your team should compare the same representative tasks across its approved Copilot surfaces and document any differences in execution or results.
For CTOs: The important shift is from a feature embedded in one editor toward an agent workflow intended to span multiple Copilot tools. That trajectory moves ownership toward the developer platform function.
Standardization should therefore cover more than installation. Teams need a shared definition of acceptable agent behavior, review requirements, and escalation paths when sessions produce different outcomes across tools.
3. Remote Task Delegation Adds Placement Decisions
What it is
Microsoft says the Agents window can delegate tasks to suitable remote hosts. Host selection can consider machine characteristics including operating system, memory, CPU, and load.
This is remote workload placement from inside the agent workflow.
Why it matters
Delegation introduces a capacity and governance decision: which hosts should receive agent tasks, and under what conditions. The release notes describe selection characteristics, but the supplied research does not establish how those characteristics should map to an enterprise’s internal policies.
Why this matters: Platform teams should define eligible host classes before broad adoption. Operating system, memory, CPU, and load are useful placement inputs, but they do not replace organizational rules for access and workload ownership.
Start with bounded tasks on approved remote hosts. Then determine whether the selection behavior aligns with your team’s capacity model and whether developers can understand where delegated work is running.
4. HydraFusion Coordinates Models and Revision
What it is
Microsoft introduced HydraFusion as a research-preview feature in VS Code 1.140. According to the release information, it coordinates model selection and workflow stages that can include critique and revision, with the stated aim of balancing speed, cost, and quality.
This is not a single-model enhancement. It is orchestration across models and workflow steps.
Why it matters
HydraFusion makes the orchestration layer an explicit part of the development experience. For enterprise teams, that creates an evaluation question: whether coordinated selection and revision produce outcomes that justify the additional workflow complexity.
Why this matters: Speed, cost, and quality are competing operating criteria. A controlled assessment should define which criterion matters for each task class before comparing HydraFusion with the team’s current workflow.
Perhaps the most consequential point is the trajectory. Agent-first development is not limited to prompting one model; it is moving toward systems that select, critique, and revise across multiple steps.
5. Enterprise Controls Govern the Default Auto Tier
What it is
Microsoft says VS Code 1.140 adds enterprise controls for configuring the default Auto tier used in model selection. The supplied research does not identify the control’s setting key or provide additional configuration details.
The policy surface is present, but the available evidence does not support a more specific configuration prescription.
Why it matters
Automatic model selection affects how teams think about cost, quality, and oversight. A configurable default gives enterprise administrators a control point, but it also requires an explicit policy decision.
What this means for IT leaders: Document whether the Auto tier is approved as a default, limited to evaluation groups, or deferred pending internal review. Do not let the editor default become the organization’s AI policy by accident.
For regulated industries, model-selection governance is non-negotiable. The immediate action is to assign ownership for the default and establish who can approve changes.
The Bottom Line
VS Code 1.140 turns Copilot agent workflows into a broader platform-management concern. Multi-folder sessions introduce isolation, the dedicated harness targets cross-tool consistency, remote delegation adds placement decisions, HydraFusion previews orchestration, and the Auto tier adds an enterprise policy surface.
What This Means for Enterprise Teams
Pilot session isolation. Test multi-folder sessions with representative repositories and review whether separate folders or worktrees clarify task ownership.
Validate harness consistency. Compare approved Copilot workflows across tools instead of accepting vendor-reported consistency without internal testing.
Constrain remote delegation. Define which remote hosts are eligible before teams route agent tasks according to operating system, memory, CPU, or load.
Separate preview evaluation. Treat HydraFusion as a research-preview assessment with explicit speed, cost, and quality criteria.
Assign model-policy ownership. Decide who controls the default Auto tier and how changes will be reviewed.
Agent-first is not a trend; it is the direction. The enterprise task is to make that direction governable.
Happy Coding!