Technology

Rider 2026.2 Ties AI Agents to the IDE, Adds Native Copilot

Rider 2026.2 connects AI agents to the IDE's own coverage, profiler, and refactoring data, and folds GitHub Copilot into the agent picker as a default option.

Rider 2026.2 connects AI agents to the IDE's coverage data, profiler snapshots, refactoring engine, and framework-specific workflows through a set of bundled skills. GitHub Copilot joins the agent picker as a natively integrated option through a JetBrains-Microsoft partnership, and users can plug in their own completion model independently of the agent provider. The release also delivers broad performance gains, XAML Hot Reload for WPF, deeper Godot and Unreal Engine support, and a long list of smaller C#, F#, and C# scripting improvements. The shape of the release signals where JetBrains sees the IDE going: a shared substrate that both human and agent developers draw on.

The pitch: IDE intelligence for agents, not the other way around

Rider 2026.2 leans into a simple idea. The IDE already knows where tests are, where the hot path is, and what the refactor engine will do. Coding agents should consume that knowledge directly instead of re-deriving it from files and terminal output. The release ships that integration as a set of bundled agent skills plus a partnership with Microsoft that puts GitHub Copilot in the agent picker out of the box, no ACP registry entry required (an active Copilot subscription is still needed for OAuth sign-in).

What the bundled skills actually do

Three skills come with the IDE. The first reads dotTrace snapshots handed to it, finds where the CPU actually went, and follows the hot path back into the code. The second runs code quality checks after every change, blocking on errors and feeding warnings back as feedback. The third reads dotCover data to figure out where new tests should go and how to format them to match the existing conventions. Other skills can be installed from inside the IDE.

If Copilot is not the right fit, the completion model can now be decoupled from the agent provider. OpenAI-compatible endpoints such as LM Studio work, and so does Mercury from Inception Labs.

The performance work is broad and concrete

Several hangs got measured and trimmed. On Windows, debugger launch for .NET apps is meaningfully faster, branch switching in Roslyn-backed solutions is generally quicker, and backend processes use less memory. For large Unreal projects opened through the generated .sln, C++ indexing is roughly half what it was in 2026.1. The C# scripting runner is a notable new addition: single-file C# scripts, repo utilities, and CI helpers no longer need a full project file.

XAML Hot Reload for WPF and a new Godot experience

WPF developers get the productivity headline of the release: XAML Hot Reload under the Rider debugger. Tweak a layout, a style, or a resource while the app is running, and the change shows up live, no rebuild, no restart. Combined with the C# Hot Reload that has been in the IDE for some time, the iteration loop on WPF UI gets noticeably tighter.

Godot development picked up a new configurable debugger, the ability to launch a project as a path or environment variable, an official Godot SDK downloader, and more accurate Autoloads resolution. On the Unreal side, UInterface navigation and Gameplay Tag usages now surface across both C++ and Blueprints, and ISPC gets the same first-class editor treatment that just landed in ReSharper C++.

Everything else that shipped

  • ReSharper engine updates: latest analysis, refactoring, and language support.
  • C++ tooling: initial support for the C++26 reflection operator, splicing, and consteval blocks, plus an evaluator that now handles dynamic allocations and exceptions.
  • C# polish: a new action to silence and restore compiler warnings, more reliable symbol imports from errors, smoother C# 14 interop.
  • Azure Functions: create, run, debug, and containerize Functions projects locally without the separate Azure Toolkit plugin.
  • GitHub pull requests: a new tool window lists, filters, reviews, votes on, and creates PRs without leaving Rider.
  • Debugger: a "preview the diff" view shows what a quick-fix or context action will change before it is applied.
  • NuGet: browsing, installed packages, and updates now have separate, more focused paths.
  • C# and F# debuggers: more predictable return values, and more reliable source generator debugging on Linux and macOS.

What this adds up to

Rider 2026.2 is the first release where the IDE treats AI agents as a first-class consumer of its own intelligence, not as a chat window bolted on the side. Whether that framing survives contact with how teams actually use these tools is an open question, but the plumbing is in place, the Copilot deal removes a meaningful onboarding step, and the performance work makes the rest of the IDE feel like it caught a second wind.