GhidraMCP
An embedded MCP server for Ghidra, exposing program data and reverse engineering functionalities.
If your browser/GitHub blocks custom URI handlers, use the web fallback: Cursor install fallback
Connect Ghidra to MCP-compatible clients
Related project: WinDbg MCP Server
✨ Features
- 30 MCP tools for reverse engineering, editing, search, and version tracking
- Structured responses with pagination and opaque cursors for large datasets
- In-memory and project-backed Version Tracking workflows for correlation, match triage, and markup
Tool Commands
- Core analysis:
analyze_rtti,decompile_code,demangle_symbol,script_guidance - Read/query:
read_symbols,read_memory_blocks,read_functions,read_data_types,read_listing,find_references,list_analysis_options,list_programs,search_memory - Modify/manage:
manage_data_types,manage_functions,manage_memory,manage_project,manage_symbols,delete_bookmark,delete_data_type,delete_function,delete_symbol - Version tracking:
manage_vt_session,run_vt_correlator,read_vt_matches,manage_vt_matches,manage_vt_markup - Utilities:
batch_operations,undo_redo,read_tool_output
🚀 Installation
- Download the latest release
zipfile from the Releases page. - In Ghidra, go to
File->Install Extensions.... - Click the
+button (Add extension) in the top right corner. - Navigate to the downloaded
zipfile and select it. - Ensure the
GhidraMCPextension is checked in the list and clickOK. - Restart Ghidra.
▶️ Usage
- Start Ghidra with the GhidraMCP extension enabled.
- Confirm the server port in Configuration.
- Point your MCP client to
http://127.0.0.1:8080/mcp(or your custom port).
[!WARNING] Script Error Dialogs: Some script-driven operations can open a Ghidra error dialog. Close the dialog before continuing, or requests may appear to hang.
[!TIP] Missing
file_name: Uselist_programsto see available programs and pass the exact name returned by that tool.
⚙️ Configuration
The GhidraMCP server can be configured through Ghidra's application-level settings:
- In Ghidra, go to Browser → Edit → Tool Options.
- In the left panel, expand Miscellaneous and select GhidraMCP HTTP Server.
- Configure the following options:
- Server Port: The port number for the MCP server (default: 8080)
- Auto-start Server: Whether to automatically start the server when Ghidra launches
- Click OK to save your settings.
🛠️ Building from Source
If you are installing from a GitHub release zip, you can skip this section. The steps below are only for building from source.
-
Clone the repository:
git clone https://github.com/themixednuts/GhidraMCP.git -
Ensure you have JDK 21 or later installed.
-
Build the project with
just:just packageTo run the same checks used by the main build CI:
just ciOr use the Gradle wrapper directly:
bash ./gradlew packageOn Windows PowerShell, use:
.\gradlew.bat packageGhidra jars are fetched automatically from the official release zip on first run.
Useful development entrypoints:
just testruns the unit suitejust test-e2eruns the end-to-end suitejust update-verification-metadatarefreshes Gradle dependency verification checksums after dependency changes
-
The installable
zipfile is written totarget/(for example,target/GhidraMCP-0.6.2.zip). Install it using the steps above.
Optional: Install Local Pre-commit Checks
To run formatting checks and full integration tests before every commit:
just install-hooks
The installed pre-commit hook runs:
just fmt-checkjust testjust test-e2e
🔌 Configuring an MCP Client
Use this server URL in your client:
http://127.0.0.1:8080/mcp(or your custom port)
Most clients use a config like:
{
"mcpServers": {
"ghidra": {
"url": "http://127.0.0.1:8080/mcp"
}
}
}
Client Setup Instructions
Config path:
- Windows:
%APPDATA%\Claude\claude_desktop_config.json - macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Linux:
~/.config/Claude/claude_desktop_config.json
Add the JSON config above, then restart Claude Desktop.
claude mcp add ghidra "http://127.0.0.1:8080/mcp" --transport http
Manual config path: ~/.cursor/mcp_settings.json
Use ~/.config/opencode/opencode.json (or project-level opencode.json):
{
"$schema": "https://opencode.ai/config.json",
"mcp": {
"ghidra": {
"type": "remote",
"url": "http://127.0.0.1:8080/mcp",
"enabled": true
}
}
}
codex mcp
Or add this to ~/.codex/config.toml:
[mcp_servers.ghidra]
url = "http://127.0.0.1:8080/mcp"
[!IMPORTANT] The default port is
8080(configurable in Ghidra: Browser → Edit → Tool Options → Miscellaneous → GhidraMCP HTTP Server). If you change the port, update your client configuration accordingly. Ghidra must be running with the extension enabled for the client to connect.
[!NOTE] Timeout Issues: If you encounter timeout problems, refer to the Ghidra timeout configuration guide.
🤝 Contributing
Contributions are welcome! Please feel free to submit pull requests or open issues.
Acknowledgements
This project is heavily inspired by and based on the work of LaurieWired. Instead of using a bridge, this plugin directly embeds the server in the plugin.
Related Servers
Scout Monitoring MCP
sponsorPut performance and error data directly in the hands of your AI assistant.
Alpha Vantage MCP Server
sponsorAccess financial market data: realtime & historical stock, ETF, options, forex, crypto, commodities, fundamentals, technical indicators, & more
FileForge
Transforms raw code into polished solutions with optimized performance and vector embeddings support.
TUUI - Tool Unitary User Interface
A desktop MCP client for tool integration and cross-vendor LLM API orchestration.
Scorecard
Access Scorecard's AI model evaluation and testing tools via a Cloudflare Workers deployment.
PyMilvus Code Generate Helper
Retrieves relevant code snippets and documents to assist in generating PyMilvus code, requiring a running Milvus instance.
pip Package README MCP Server
Fetch READMEs, metadata, and search for Python packages on PyPI.
MCP Memory Visualizer
Graph visualization tools for exploring and analyzing Claude's memory data.
MCP Analytics with GitHub OAuth
A remote MCP server with GitHub OAuth authentication and built-in analytics tracking.
CrowdCent MCP Server
Integrates with the CrowdCent Challenge API, allowing AI assistants to manage prediction challenges, datasets, and submissions.
Databutton
An MCP server for initial app planning and creating a good starting point for an app.
302AI Sandbox MCP Server
A code sandbox for AI assistants to safely execute arbitrary code. Requires a 302AI API key for authentication.