RustRover 2026.3 Is Out: JetBrains Air, Clearer Debugger Views, and Agents That Debug With You
The JetBrains Blog · 软件开发
Fixing a bug starts with understanding what your code is doing at runtime: reading the program state, following the call stack, or tracking down a slow function. RustRover 2026.3 makes that work easier to follow, whether you do it yourself or with an AI agent. It also ships with JetBrains Air , a new layer for running coding agents and reviewing their work in the IDE. Here’s a quick look at the highlights. For full details, settings, and screenshots, see What’s New in RustRover 2026.3 . Download RustRover 2026.3 Run and review your agents’ work in JetBrains Air JetBrains Air works with the agents you already use, including Claude Agent, Codex, Gemini, GitHub Copilot, Junie, OpenCode, and other ACP-compatible agents. Your existing subscriptions and API keys carry over, with no JetBrains AI subscription required. Junie Lite gives you a free agent with just a JetBrains Account, and Junie can also work with models running locally through Ollama or LM Studio. When an agent makes changes, you can review them with the surrounding code in view and anchor notes to specific lines in the diff, so the agent knows exactly what to fix. Read debugger views at a glance Long, fully qualified Rust names can crowd the debugger. RustRover now shortens type names in the Variables and Watches views and symbol names in the Frames view, so the program state and the call stack are easier to scan. Shortening is on by default, and you can turn it off in the settings. Go from profiler results to the code behind them When profiler results point you to a hot spot, Jump to Source now opens the Rust function behind it instead of reporting Declaration not found . Function names in the results are readable instead of mangled. Both improvements work with perf on Linux and DTrace on macOS. Let agents use the debugger RustRover now provides a debugger MCP server and a rustrover-debugger skill that teaches agents how to use it. Instead of stepping through the code yourself, you describe the problem in n