abhishekrn44

Jirrafe — AI skill for Claude Code

AI community

Code graph of Java/Kotlin projects, including the code inside internal jars, for AI coding agents.

How to install Jirrafe

This entry records only its repository, not the path inside it, so there is no exact command to give. Open abhishekrn44/jirrafe and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Jirrafe does

Code graph of Java/Kotlin projects, including the code inside internal jars, for AI coding agents. Ships as an agent skill and an MCP server (Claude Code, Cursor, Copilot, Windsurf, Codex): cited answers to "how does X work", "what calls Y", "what breaks if I change Z" in one or two calls instead of a grep session.

Alternatives in AI

Token Monitor — Local-first desktop widget for tracking token usage, costs, and limits across 32+ AI coding tools—including Cl 1.8k ★
Nocturne Memory — A lightweight, rollbackable, and visual Long-Term Memory Server for MCP Agents 1.3k ★
Internal Safety Collapse — We built an adversarial codespace setup 1.2k ★

README

jirrafe

A code graph of any Gradle or Maven Java or Kotlin project, **including the code inside its internal jars**, that an AI coding agent queries instead of grepping. It ships as a Claude Code skill and as an MCP server for GitHub Copilot, Cursor, Windsurf and any other MCP client. An agent asks "how does X work", "what calls Y" or "what breaks if I change Z" and gets the flows, the classes and methods, their callers and their source, every one cited as `file:line`, in a few calls instead of a grep session.

**76% fewer tokens per question** than grepping for the same answer, in one or two tool calls instead of a dozen, measured on repositories the model had not memorised ([What to expect](#what-to-expect)).

Requirements: JDK 17 or 21; the project builds with Gradle 7.6+ or Maven 3.9. Kotlin modules are indexed from bytecode.

Why

In enterprise Java most real logic lives in internal jars pulled from a private repository. Every source-based assistant goes blind at that boundary and guesses. jirrafe reads those jars with ASM and their sources jars with javac, resolves calls with real Java semantics, models Spring wiring, precomputes one flow per entry point, and hands an agent a cited map instead of a search box. The compiler is what makes the answers exact, and it needs your build, which you already have.

Install

Until the first release is published, build from source. macOS and Linux:

git clone https://github.com/abhishekrn44/jirrafe.git && cd jirrafe
./gradlew :jirrafe-cli:installDist
export PATH="$PWD/jirrafe-cli/build/install/jirrafe/bin:$PATH"   # add to ~/.zshrc or ~/.bashrc to keep it
jirrafe --help

Windows (PowerShell):

git clone https://github.com/abhishekrn44/jirrafe.git; cd jirrafe
.\gradlew.bat :jirrafe-cli:installDist
$env:PATH = "$PWD\jirrafe-cli\build\install\jirrafe\bin;$env:PATH"
jirrafe --help

That PATH lasts for the session; to keep it, add the `bin` directory under System Properties -> Environment Vari