Microsoft shipped Visual Studio Code 1.142 on October 14, 2026, and the release notes point to the clearest signal yet that the company sees its flagship code editor as an agent-first workspace rather than a text editor with extensions bolted on. The update, detailed on the official VS Code 1.142 release notes, introduces interactive canvases for collaborating with coding agents, lets developers carry chat context across folders and Git worktrees, and expands enterprise controls over what autonomous agents are allowed to touch. It lands alongside a wave of Python tooling fixes and arrives the same week the Go ecosystem pushed its own agent-adjacent tooling forward with gopls 0.24.0 and VS Code Go 0.56.0.
None of this happens in a vacuum. VS Code remains the most-used code editor among professional developers, and Microsoft’s decision to keep layering agent infrastructure directly into the editor, rather than treating it as a bolt-on extension, is a direct response to competitive pressure from Cursor, Windsurf, Zed, and JetBrains’ AI Assistant. The vs code 1.142 update is also a window into where the broader coding-tools market is heading in late 2026: fewer standalone AI chat sidebars, more persistent, governed, sandboxed agent sessions that enterprises can actually sign off on.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
What Changed in the VS Code 1.142 Update
The headline feature in VS Code 1.142 is the introduction of interactive canvases for agent collaboration, a preview surface that gives developers a dedicated space to work alongside coding agents rather than reading agent output as a scrolling chat log. According to Microsoft’s release documentation, the canvas sits next to the editor and is designed to make multi-step agent work easier to follow and interrupt. The release also adds expanded enterprise controls for agent permissions and introduces improvements to agent sandboxing, two areas that have become central to how IT and security teams decide whether to let AI agents touch production code at all.
Beyond the agent-facing changes, the release notes list improvements to model selection inside the editor’s AI-assisted tools, expanded chat-management features for organizing ongoing development conversations, and a round of fixes to the Markdown editor. Taken together, the changes read less like a single flagship feature and more like Microsoft methodically hardening every seam where a developer might hand off work to an autonomous agent and need to pull it back.
Persistent Chat Context Across Folders and Git Worktrees
One of the more practical changes in VS Code 1.142 addresses a complaint that has followed agentic coding tools since they first shipped: context resets the moment you switch folders or check out a different Git worktree. The 1.142 release lets developers move an active chat session between folders and worktrees while preserving the conversation’s context, a workflow detail that matters more than it sounds. Developers increasingly juggle multiple Git worktrees when working on parallel branches or reviewing a colleague’s pull request without disturbing their own working copy, and losing agent context every time they switch has been a persistent source of friction.
This single fix illustrates the broader pattern across the release: Microsoft is not chasing a flashier agent demo so much as removing the rough edges that keep agent-assisted coding from fitting into how teams already use Git. That is a sharply different posture than the “look what the agent can build from a blank prompt” marketing that defined 2024 and 2025 agent launches, and it lines up with how Tech Insider has covered the broader shift toward developers spending more of their week reviewing AI-generated code rather than writing it from scratch.
Python Tooling Gets Real Fixes, Not Just Features
Python remains one of the most widely used languages inside VS Code, and the 1.142 release spends real effort on it without adding anything flashy. The Python Environments Extension received fixes for environment discovery, meaning the editor should more reliably detect virtual environments, conda environments, and interpreters that previously required manual configuration. Interpreter-selection behavior was also improved, and the extension picked up fixes for terminal activation, addressing a recurring annoyance where the correct interpreter would load in the editor but the integrated terminal would default to the wrong one.
These are not the kind of changes that generate headlines, but they are the kind that show up in day-to-day developer complaints on forums and in GitHub issue trackers. For a release otherwise centered on agent infrastructure, the Python fixes are a reminder that Microsoft still has to maintain VS Code as a working tool for the millions of developers who have no interest in agent canvases and just want their interpreter to load correctly.
Enterprise Governance: Agent Permissions and Sandboxing
The expanded controls over agent permissions and sandboxing in VS Code 1.142 point to where Microsoft believes the real adoption bottleneck sits: not with individual developers, who have largely embraced AI-assisted coding, but with the security and compliance teams that have to approve what those agents can do inside a corporate codebase. Sandboxing controls let organizations constrain what an agent can read, write, or execute during a session, while expanded permission settings give administrators finer-grained ability to decide which agent capabilities are available by default versus which require explicit approval.
This mirrors a tension Tech Insider has tracked across the coding-tools space this year: benchmarks keep improving on how well AI coding agents pass functional tests, but separate security research has repeatedly flagged weaker results on secure-coding practices. A widely cited 2026 benchmark found AI coding agents passing the large majority of functional tests while only a quarter of the generated code met secure-coding standards, which is precisely the gap that sandboxing and permission controls are designed to manage rather than solve outright.
Why Governance Features Are Becoming a Sales Pitch
Enterprise buyers evaluating coding tools in 2026 are no longer just comparing autocomplete quality. They are asking whether an agent’s actions are logged, whether its permissions can be scoped per repository, and whether a sandboxed session can be audited after the fact. Microsoft folding these controls directly into VS Code, rather than leaving them to third-party extensions, is as much a procurement argument aimed at CISOs as it is a developer-experience improvement.
Markdown Editor Improvements Round Out the Release
VS Code 1.142 also ships a set of improvements to the built-in Markdown editor, a less glamorous but widely used part of the tool given how much documentation, README content, and AI-agent planning output now lives in Markdown files. As agents increasingly write project plans, changelogs, and documentation as part of their output, a smoother Markdown editing experience is less of a side feature and more of a direct dependency for how teams review and merge that output.
The Ecosystem Moves Too: gopls 0.24.0 and VS Code Go 0.56.0
The VS Code 1.142 release did not land in isolation. The same October window saw the Go language server, gopls, reach version 0.24.0, while the VS Code Go extension advanced to version 0.56.0, adding interactive refactorings for struct tags and interface implementations, according to the VS Code Go release notes on GitHub. Refactoring tools that understand struct tags and interface implementations are a meaningful quality-of-life jump for Go developers working in large codebases, where manually keeping struct tags and interface satisfaction in sync across dozens of files is a routine source of tedious, error-prone edits.
The timing is not coincidental. Language-specific tooling teams inside and adjacent to Microsoft tend to ship updates in clusters around major VS Code releases, and the Go tooling bump gives Microsoft another concrete example to point to when arguing that the core editor update and its language ecosystem are improving in lockstep, not competing for engineering attention.
Why Microsoft Is Racing to Make VS Code Agent-Native
The strategic logic behind VS Code 1.142 is straightforward once you place it next to the rest of the coding-tools market. Cursor, built as a VS Code fork with AI-native workflows from day one, has pulled meaningful share away from vanilla VS Code usage specifically because it treated agentic editing as the core product rather than an add-on. Microsoft’s answer has been to close that gap feature by feature inside the original editor: agent mode, now agent canvases, persistent chat context, and governance controls that a VS Code fork without Microsoft’s enterprise relationships would struggle to match.
That calculation also explains why GitLab, a company with no code editor of its own, chose to integrate directly with next-generation models inside its Duo agent platform rather than build a competing editor, a move Tech Insider covered when GitLab Duo added Gemini 4 Argon support to its agent platform. The editor layer is increasingly a distribution channel for whichever AI backend wins trust with enterprise buyers, and Microsoft wants VS Code to remain the default surface where that trust gets built.
Competitive Landscape: VS Code vs Cursor vs JetBrains vs Zed vs Windsurf
The 2025 Stack Overflow Developer Survey offers the clearest independent snapshot of where developers actually stand on editor choice, even though its figures reflect reported usage across a broad respondent base rather than strict commercial market share. The numbers below, drawn from that survey and from JetBrains’ own 2025 ecosystem research, show VS Code still comfortably ahead of every AI-native challenger, even as those challengers post real usage numbers of their own.
| Editor / IDE | 2025 Reported Usage | Primary AI Approach | Governance Controls |
|---|---|---|---|
| Visual Studio Code | 75.9% (Stack Overflow 2025 survey) | Built-in agent mode, agent canvases (1.142) | Enterprise agent permissions, sandboxing |
| JetBrains IntelliJ IDEA | 27.1% (Stack Overflow 2025 survey) | JetBrains AI Assistant | Enterprise license tiers, IDE-level policy |
| Cursor | 17.9% (Stack Overflow 2025 survey) | AI-native fork of VS Code | Team/enterprise privacy modes |
| PyCharm | 15.0% (Stack Overflow 2025 survey) | JetBrains AI Assistant | Enterprise license tiers, IDE-level policy |
| Zed | 7.3% (Stack Overflow 2025 survey) | Collaborative, performance-first AI editing | Early-stage team controls |
| Windsurf | 4.9% (Stack Overflow 2025 survey) | Agentic “flows” editing model | Enterprise plan controls |
Feature Parity Is Closing, Pricing Clarity Is Not
What the table does not show is pricing, and that is deliberate: none of the specific October 2026 pricing figures for GitHub Copilot tiers, Cursor, Windsurf, or Zed could be independently verified for this article, so they are left out rather than guessed at. What is verifiable is the direction of travel. Every major AI-assisted editor now ships some version of agent mode, persistent context, and enterprise governance controls. The competitive battle in late 2026 is less about who has an agent feature and more about whose governance model a security team will actually approve for use on proprietary code.
The Adoption Numbers Behind the Agentic Coding Shift
Separate from editor choice, the rate at which developers are adopting AI coding assistance at all has climbed sharply. The 2025 Stack Overflow Developer Survey found that the large majority of developers now use or plan to use AI tools in their workflow, up from 76% to 84% since 2024. JetBrains’ State of Developer Ecosystem 2025 survey, based on more than 24,000 developers across 194 countries, found a similar pattern from an independent respondent pool, as detailed in the JetBrains State of Developer Ecosystem 2025 report.
| Metric | Figure | Source |
|---|---|---|
| Developers using or planning to use AI tools | 84% (up from 76% in 2024) | Stack Overflow Developer Survey 2025 |
| Developers currently using AI agents | 31% | Stack Overflow Developer Survey 2025 |
| Developers using AI agents daily at work | 14.1% | Stack Overflow Developer Survey 2025 |
| Developers regularly using AI tools for coding | 85% | JetBrains State of Developer Ecosystem 2025 (24,534 respondents, 194 countries) |
| Developers relying on at least one AI coding assistant, agent, or editor | 62% | JetBrains State of Developer Ecosystem 2025 |
| Developers who had not adopted AI tools in daily work | 15% | JetBrains State of Developer Ecosystem 2025 |
Read together, the two surveys use different populations and definitions, so the individual figures should not be averaged into a single number. But the direction each one points is consistent: a clear majority of developers have moved past treating AI coding assistance as a novelty, which is exactly the environment in which a feature like VS Code’s agent canvases stops being a demo and starts being daily-use infrastructure.
Historical Context: From IntelliSense to Autonomous Agents
Visual Studio Code launched in 2015 as a lightweight, fast alternative to heavier IDEs, built around IntelliSense autocomplete and a growing extension marketplace. For most of the following decade, its evolution was incremental: better language servers, a built-in terminal, remote development support, and a steady stream of editor polish. The shift toward AI assistance began in earnest with GitHub Copilot’s integration as an extension, which turned VS Code into the default on-ramp for AI-assisted autocomplete well before “agent mode” entered the vocabulary.
What VS Code 1.142 represents is the next stage in that same arc: moving from autocomplete, to chat-based assistance, to semi-autonomous agents that can carry out multi-step tasks across a codebase, and now to agents with persistent memory, dedicated collaboration surfaces, and enterprise-grade governance. Each stage has compressed the previous one’s novelty into table-stakes functionality, which is why a release focused on agent canvases and sandboxing reads as expected progress rather than a surprise pivot.
Security Stakes: What Sandboxing Actually Buys Enterprises
Sandboxing an AI coding agent means constraining the file system, network, and execution permissions available to it during a session, so that even if an agent is manipulated into taking a harmful action through a bad prompt or a malicious instruction hidden in a file, the blast radius is limited. That is a meaningfully different security model than simply trusting the agent’s output after the fact, and it reflects growing enterprise awareness that agents operating with broad repository access are a real attack surface, not a theoretical one.
Teams evaluating VS Code 1.142’s sandboxing controls should treat them as a floor, not a ceiling. A sandbox limits what an agent can do in a single session; it does not replace code review, secret scanning, or dependency auditing. A setup like the one below is a common enterprise baseline for constraining what an agent session can reach before it ever gets near production credentials.
{
"chat.agent.sandboxing": "strict",
"chat.agent.permissions.filesystem": "workspace-only",
"chat.agent.permissions.network": "deny",
"chat.agent.permissions.terminal": "prompt"
}Market Impact on Developer Tooling Budgets
For engineering leaders, VS Code 1.142’s governance features change the calculus on tooling spend. Instead of paying for a separate AI governance layer bolted onto whichever editor a team already uses, organizations get baseline permission and sandboxing controls for free inside the editor most of their developers already run. That does not eliminate the market for dedicated AI code-review and security tooling, a space Tech Insider has tracked as agents increasingly generate code that passes functional tests but still carries security gaps, but it does shift where enterprises draw the line between “editor feature” and “paid add-on.”
It also raises the competitive bar for rivals. A security or governance feature that used to be a selling point for a premium AI coding platform is now a free update inside the world’s most-used editor, which puts pressure on Cursor, Windsurf, and JetBrains to either match the governance baseline or differentiate more sharply on raw agent capability and speed.
Predictions: Where Agentic Code Editors Go Next
- Expect VS Code’s agent canvases to move from preview to default-on within one or two subsequent releases, following the same trajectory agent mode itself took.
- Enterprise governance controls, not raw model capability, will become the primary battleground among code editors through 2027, as functional benchmarks saturate and security gaps become the differentiator buyers actually ask about.
- More language-server teams will time feature releases to coincide with major VS Code updates, following the pattern set by gopls 0.24.0 and VS Code Go 0.56.0 this cycle.
- Pricing transparency will become a competitive issue in its own right, as buyers increasingly compare governance and agent features across tools without clear, comparable published pricing.
- Persistent, cross-context agent memory, the kind introduced for Git worktrees in 1.142, will expand to cover cross-repository and cross-session memory as the next logical frontier.
What Developers and Teams Should Do Now
Developers running VS Code should update to 1.142 to pick up the Python environment and interpreter fixes alone, which resolve real daily friction independent of any interest in agent features. Teams already using agent mode should review the new sandboxing and permission settings before their next sprint, since the defaults may be more permissive than what a security review would prefer. Engineering leaders evaluating editor standardization should treat this release as evidence that the gap between VS Code and AI-native forks like Cursor is narrowing on governance, even if raw agent capability still varies by tool.
For teams still deciding between a standardized editor and letting developers choose freely, the more relevant comparison today is not “VS Code or Cursor” in isolation, but how each tool’s governance model fits the organization’s existing security posture, a question that Tech Insider has previously benchmarked across diff-based versus full-rewrite agent editing behavior in a direct Cursor versus VS Code codebase benchmark.
FAQ
When did VS Code 1.142 release?
Microsoft’s release notes list the stable VS Code 1.142 update as shipping October 14, 2026.
What is the biggest new feature in VS Code 1.142?
The headline addition is interactive agent collaboration canvases, a preview surface for working alongside coding agents, alongside persistent chat context across folders and Git worktrees and expanded enterprise agent permission and sandboxing controls.
Does VS Code 1.142 fix Python issues?
Yes. The Python Environments Extension received fixes for environment discovery, interpreter selection, and terminal activation in this release.
How does VS Code’s agent sandboxing work?
Sandboxing controls let administrators and developers constrain what an agent session can read, write, or execute, limiting the impact if an agent is misused or manipulated during a session rather than relying solely on reviewing its output afterward.
Is VS Code still more popular than Cursor and other AI-native editors?
Yes, by a wide margin on reported usage. The 2025 Stack Overflow Developer Survey found 75.9% of respondents using VS Code versus 17.9% for Cursor, though these figures are not mutually exclusive since developers can and do use multiple editors.
What is gopls and why does its update matter for VS Code users?
Gopls is the official Go language server. Its 0.24.0 update, paired with VS Code Go 0.56.0, adds interactive refactorings for struct tags and interface implementations, directly improving the Go development experience inside VS Code the same month as the 1.142 release.
Should enterprises change their AI coding policy because of this release?
Enterprises already permitting AI coding agents should review the new permission and sandboxing defaults in 1.142, since baseline settings may be more permissive than internal security policy requires, and should not treat sandboxing as a replacement for code review or secret scanning.
Is GitHub Copilot part of the VS Code 1.142 update?
The 1.142 release notes focus on editor-level agent infrastructure, chat management, and model-selection improvements that affect how AI assistance behaves inside the editor, rather than announcing a specific new Copilot product tier.
