kawaiidra

Kawaiidra MCP is a Ghidra-based MCP server designed to power Claude Code and other MCP-compatible AI assistants with binary analysis capabilities. It exposes a range of analysis tools that let you import binaries, decompile functions, inspect disassembly, locate cross-references, extract data structures, and export results as JSON for downstream processing. The server can operate in headless mode or be connected to a running Ghidra GUI via ghidra_bridge for real-time context tracking. You can run it via a Python-based MCP setup and connect Claude Code to the configured mcpServers entry. Typical workflows include loading a binary into a Ghidra project, analyzing functions to obtain decompiled C code, browsing cross-references, and exporting a structured report for integration with security research or malware reverse engineering tasks.

Prerequisites:\n- Python 3.10+ (recommended)\n- Ghidra 11.x or 12.x installed on your system\n- MCP Python package (pip install mcp) if needed for your environment\n- Java (for JPype-based acceleration, optional but recommended)\n\n1) Install Ghidra:\n- Option A: Use Homebrew (macOS/Linux): brew install ghidra\n- Option B: Manual: Download from ghidra-sre.org and extract it.\n\n2) Prepare Python environment:\nbash\npython -m venv venv\nsource venv/bin/activate # Linux/macOS\nvenv\Scripts\activate # Windows\npip install -r requirements.txt\n\n\n3) Install JPype (optional for speed):\nbash\npip install JPype1\n\n4) Ensure Ghidra is discoverable by the server:\n- If not auto-detected, set GHIDRA_INSTALL_DIR to the Ghidra installation path:\nbash\nexport GHIDRA_INSTALL_DIR=/path/to/ghidra # Linux/macOS\nset GHIDRA_INSTALL_DIR=C:\path\to\ghidra # Windows\n\n5) Run the MCP server:\nbash\n# From the kawaiidra-mcp directory\npython run_server.py\n\n6) Connect Claude Code to the MCP server using the provided mcp.json configuration.\n

Tips and common issues:\n- Ensure GHIDRA_INSTALL_DIR points to a valid Ghidra installation; some setups auto-detect, but explicit setting helps.\n- If you enable JPype, Java 17+ must be available on your system. Check java -version.\n- Binaries should be placed in a project directory (e.g., binaries/) or referenced via absolute paths.\n- Use the bridge_status tool to verify if the JPype bridge is active for faster operation.\n- When exporting analyses, use export_analysis to JSON to integrate with CI pipelines or other tooling.\n- If the server cannot locate Ghidra, ensure Ghidra's native libraries are accessible (LD_LIBRARY_PATH on Linux, PATH on Windows).