MCPSERV.CLUB
jtang613

GhidrAssistMCP

MCP Server

AI‑powered reverse engineering via Ghidra’s MCP server

Active(74)
323stars
1views
Updated 12 days ago

About

A Ghidra extension that implements the Model Context Protocol, exposing 31 built‑in analysis tools over a standard API. It enables AI assistants and external scripts to query program, function, and data information in real time.

Capabilities

Resources
Access data sources
Tools
Execute functions
Prompts
Pre-built templates
Sampling
AI model interactions

Screenshot

GhidrAssistMCP is a Ghidra extension that turns the powerful reverse‑engineering environment into a first‑class AI‑ready platform. By implementing the Model Context Protocol (MCP), it exposes Ghidra’s rich analysis engine through a standardized, HTTP‑based API that any AI assistant—such as Claude or custom scripts—can consume. This bridges the gap between human‑driven reverse engineering and automated, AI‑augmented analysis workflows.

The server provides a comprehensive set of 31 built‑in tools that cover program, function, data, and namespace discovery. These tools are designed to be context‑aware: they automatically use the current cursor position or active function, allowing an AI assistant to query “What’s happening at this address?” or “Show me all strings referenced by the current function” without manual parameter passing. The tool set includes high‑level queries like , , and , as well as more granular calls such as or . Because each tool is independently enableable, developers can tailor the server’s surface area to match their workflow or security policy.

A key value proposition is the dynamic tool management and persistent configuration. Through a lightweight UI, users can enable or disable individual tools on the fly and save those preferences across sessions. Real‑time logging of every MCP request and response gives complete auditability, which is essential in regulated or collaborative environments. The server runs on a configurable host and port (defaulting to ), making it easy to integrate into existing network topologies or containerized deployments.

In practice, GhidrAssistMCP empowers a range of use cases: an AI assistant can walk through a binary, automatically generate documentation, or suggest refactoring opportunities; automated analysis pipelines can trigger Ghidra’s powerful decompilation and cross‑reference features without manual intervention; researchers building custom plugins can leverage the same MCP interface to harness Ghidra’s data in external ML models. Because the protocol is standardized, any future AI tool that supports MCP can immediately tap into Ghidra’s capabilities, ensuring long‑term interoperability.

Overall, GhidrAssistMCP turns Ghidra from a standalone desktop application into a versatile AI‑first service. It offers developers a clean, extensible interface to Ghidra’s analytical depth while preserving the familiar visual workflow. This combination of rich tooling, context awareness, and protocol‑level integration makes it a standout solution for teams seeking to augment reverse engineering with intelligent automation.