JetBrains Introduces Plugin for Modern Go Code with AI Agents Junie and Claude Code

JetBrains has launched a new plugin that enhances AI agents Junie and Claude Code, enabling them to write modern Go code. This plugin integrates guidelines to ensure code generation utilizes the latest Go features as defined in the go.mod file of the user's project.
Key Features
- The plugin updates AI code generation to align with the current version of Go, mitigating issues of outdated code suggestions. For example, it replaces manual loops with the
slices.Contains()method for checking elements in a slice, a feature introduced in Go 1.21. - Data cutoff and frequency bias are managed by the plugin. Data cutoff refers to the training data not covering the newest Go releases, while frequency bias leads AI models to suggest commonly used but outdated code patterns.
Integration Details
- For Junie: Version 2xx.620.xx or newer automatically includes the plugin's guidelines. Users of older versions need to update via Settings → Plugins → Installed. The feature can be disabled in Settings → Tools → Junie → Project Settings → Go.
- For Claude Code: Install the plugin by adding the repository to the marketplace using the command:
/plugin marketplace add JetBrains/go-modern-guidelines. Then, install it with/plugin install modern-go-guidelines. Activate the guidelines with/use-modern-go.
This tool is particularly relevant for developers aiming to ensure their codebase remains updated with contemporary Go standards and practices, effectively preventing the proliferation of obsolete code patterns.
📖 Read the full source: HN AI Agents
👀 See Also

Swarm Leak Detector: Free Tool to Scan for Exposed API Keys in OpenClaw Configs
A developer released swarm-leak-detector, a zero-dependency MIT-licensed tool that scans for over 21 credential patterns (OpenAI, Anthropic, OpenRouter, Stripe, etc.) in plaintext JSON config files. Run it with npx swarm-leak-detector scan ~/.clawdbot/ to check for leaks in about 30 seconds.

LLM Agent Builds Complete Godot 4 Dungeon Crawler Using Visual Feedback
A developer connected an LLM agent to Godot 4 using an MCP tool and gave it a single prompt to build a dungeon crawler FPS. The agent created a complete prototype with 3 rooms, lighting, combat, enemies, and progression by running the game, taking screenshots, and fixing visual issues.

Stop Re-Teaching Claude Code Every Session: Use a Persistent Config
A Reddit user explains how they saved 20 minutes per session by writing a persistent config for Claude Code, eliminating repetitive steering and achieving 33% faster completions.

Mymir: An Open-Source Project Graph for Coding Agents via MCP
Mymir gives agents a graph-based project map with dependencies, decisions, acceptance criteria, and previous execution notes, delivered via MCP to avoid re-explaining state across sessions.