Visual Studio 2026 18.9.0: Native AI Code Review, Thinking Effort Controls, and the Enterprise Answer to Cursor

Adjustable thinking effort controls

Supported Copilot models in Visual Studio 18.9 now include a thinking effort selector with low, medium, and high levels . The control adjusts how deeply the model reasons before generating a response . Higher effort improves answers for complex problems but consumes more tokens and AI Credits under the usage-based billing model that began in June 2026 .

Microsoft's documentation explains the levels: "Low gives quick responses with minimal reasoning, best for straightforward questions and code suggestions. Medium balances reasoning depth and response speed. High gives deeper reasoning for complex problem-solving" . The specific levels available depend on the model selected . This feature, first introduced in VS Code 1.113 in March 2026, gives developers a direct knob to balance response quality against cost and latency — a critical capability as AI usage shifts to token-based billing .

Redesigned model picker with cost and capability info

Visual Studio 18.9 overhauls the model picker. Developers can now pin favorite models to keep them at the top, collapse models they don't frequently use, and see cost, capability, and estimated per-turn credit consumption directly in the dropdown . A companion Copilot Usage window has been refreshed to show token-level billing rather than per-request counts, with real-time usage metrics and proactive alerts when approaching limits .

This redesign reflects the post-June 2026 reality of usage-based billing. As one analysis noted, "what consumes AI Credits is everything categorized as an agentic interaction: Copilot Chat conversations, Agent mode sessions, the cloud agent that creates GitHub issues and pull requests autonomously, and code review" . By surfacing cost information directly in the model picker, Microsoft aims to prevent billing surprises while giving developers the flexibility to choose the right model for the task.

Organization-wide custom agents

GitHub organization owners can now add custom agents across their entire organization, sharing team-specific agent instructions and tools across repositories . This builds on the custom agents framework introduced earlier in 2026, where agents are defined in .agent.md files stored in .github/agents/ in a repository or in a user's profile . Built-in .NET and Azure skills authored by Microsoft teams are also available in the tool picker .

Organization-wide custom instructions, introduced in the July 2026 update, allow GitHub organization owners to tailor Copilot's responses across all team members automatically . This enterprise governance capability is something that newer AI-native tools like Cursor and Claude Code currently lack — and it directly targets Microsoft's existing enterprise customer base.

How 18.9.0 fits into the competitive landscape

Microsoft's AI development platform strategy

The 18.9 release is the strongest signal yet of Microsoft's strategy to embed AI natively into Visual Studio's platform rather than offering it as a bolt-on. The July 2026 update introduced a new Agent (Preview) built on the GitHub Copilot SDK, unifying the agent experience across CLI, VS Code, and Visual Studio . The shift to token-based billing, cost transparency, and model choice all point to Microsoft treating AI as a first-class metered resource within the IDE.

Direct competition with Cursor and Claude Code

Visual Studio 18.9 directly counters the momentum of newer AI-native coding tools:

• Cursor popularized agentic editing with inline diffs, multi-file edits, and model flexibility. Visual Studio 18.9 matches this with Copilot Edits across files, the Git agent for pre-commit review, and full model picker flexibility .
• Claude Code (Anthropic's terminal-based coding agent) and third-party extensions like CodeRabbit already offer AI code review of uncommitted changes . By building the Git agent natively into the IDE rather than requiring extensions, Microsoft aims to close the gap without developers leaving Visual Studio.
• Custom agents at org scale addresses a differentiator that Cursor and Claude Code lack: enterprise-grade governance for AI tooling. Organization-wide custom agents allow IT and platform teams to enforce coding standards and provide curated AI personas across the entire company.

The broader theme is that Visual Studio 2026 is no longer a traditional IDE with a Copilot sidebar — it is repositioning as an AI-integrated development platform with agentic workflows, fine-grained model control, cost governance, and team-scale customization, all designed to keep enterprise teams inside the Microsoft ecosystem as the AI coding tool landscape fragments.