Skip to content

knifeyspoony/ghidra-copilot

Repository files navigation

Ghidra Copilot

AI-assisted companion for Ghidra. It embeds a chat panel that can answer questions about your current program, suggest next steps, and run small analysis tools directly from the conversation.

What It Does

  • Chat inside Ghidra that stays in context with the open program.
  • Tool calls for symbol search, data extraction, and refactoring helpers.
  • Works with OpenAI, Azure OpenAI, Anthropic, and Ollama through Spring AI.

Screenshots

  • Chat panel: screenshot coming soon
  • Tool results: screenshot coming soon

Requirements

  • Ghidra 11.4.2 or compatible
  • Java 17+
  • Network access or an on-prem model endpoint supported by Spring AI

Install

  1. Set GHIDRA_INSTALL_DIR to your Ghidra path.
    • Linux/WSL: export GHIDRA_INSTALL_DIR=/absolute/path/to/ghidra_11.4.2_PUBLIC
    • Windows PowerShell: $env:GHIDRA_INSTALL_DIR = "D:\\tools\\ghidra_11.4.2_PUBLIC"
  2. Build the extension:
    • Linux/WSL: ./gradlew buildExtension
    • Windows PowerShell: .\\gradlew buildExtension
  3. Install the generated zip from dist/ through Ghidra’s Install Extensions dialog.

Configure Models

Inside Ghidra, open Tool Options > Ghidra Copilot to pick a provider and enter the API key/base URL/model names. Settings are stored in your Ghidra user directory and reused on next launch:

  • OpenAI: API key (optional base URL/model override)
  • Azure OpenAI: API key, endpoint, deployment, optional model
  • Anthropic: API key, model
  • Ollama: base URL, model

Use It

  1. Open Ghidra and enable Ghidra Copilot in the tool window.
  2. Ask questions about the current program; tool calls run automatically when needed.
  3. Use the panel settings to switch model providers or keys.

About

Provides a chat-based, LLM-assisted reverse engineering experience within Ghidra

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages