docs: Rewrite README for clarity and current features
Some checks are pending
Build Ghidra Plugin / build (push) Waiting to run
Some checks are pending
Build Ghidra Plugin / build (push) Waiting to run
- Punchy hero section with terminal-style demo - Feature table showing all 64 tools across 12 categories - Docker quickstart as primary path (easiest) - Clear usage patterns: current instance, Docker workflow, pagination - Concise tool reference grouped by category - Architecture notes explaining AI-agent design decisions - Removed outdated v2.1 references and verbose API examples
This commit is contained in:
parent
41bd8445e9
commit
04f3011413
753
README.md
753
README.md
@ -1,601 +1,302 @@
|
|||||||
|
# GhydraMCP
|
||||||
|
|
||||||
|
**AI-native reverse engineering.** Give Claude (or any MCP client) direct access to Ghidra's analysis engine.
|
||||||
|
|
||||||
|
```
|
||||||
|
┌─────────────────────────────────────────────────────────────────────────────┐
|
||||||
|
│ "Analyze the authentication bypass in this firmware" │
|
||||||
|
│ │
|
||||||
|
│ Claude: I'll decompile the auth functions and trace the validation logic. │
|
||||||
|
│ │
|
||||||
|
│ [functions_list grep="auth|login|verify"] │
|
||||||
|
│ [functions_decompile name="verify_password"] │
|
||||||
|
│ [xrefs_list to_addr="0x0040156c"] │
|
||||||
|
│ [analysis_get_dataflow address="0x00401234" direction="backward"] │
|
||||||
|
│ │
|
||||||
|
│ Found it. The password check at 0x401580 compares against a hardcoded │
|
||||||
|
│ hash, but there's a debug backdoor at 0x401590 that bypasses validation │
|
||||||
|
│ when the username starts with "debug_". Let me show you the call graph... │
|
||||||
|
└─────────────────────────────────────────────────────────────────────────────┘
|
||||||
|
```
|
||||||
|
|
||||||
[](https://www.apache.org/licenses/LICENSE-2.0)
|
[](https://www.apache.org/licenses/LICENSE-2.0)
|
||||||
[](https://github.com/starsong-consulting/GhydraMCP/releases)
|
|
||||||
[](https://github.com/starsong-consulting/GhydraMCP/blob/main/GHIDRA_HTTP_API.md)
|
|
||||||
[](https://github.com/starsong-consulting/GhydraMCP/stargazers)
|
|
||||||
[](https://github.com/starsong-consulting/GhydraMCP/network/members)
|
|
||||||
[](https://github.com/starsong-consulting/GhydraMCP/graphs/contributors)
|
|
||||||
[](https://github.com/starsong-consulting/GhydraMCP/actions/workflows/build.yml)
|
|
||||||
|
|
||||||
# GhydraMCP v2.1
|
## What You Get
|
||||||
|
|
||||||
GhydraMCP is a powerful bridge between [Ghidra](https://ghidra-sre.org/) and AI assistants that enables comprehensive AI-assisted reverse engineering through the [Model Context Protocol (MCP)](https://github.com/modelcontextprotocol/mcp).
|
**64 MCP tools** across 12 categories:
|
||||||
|
|
||||||
|
| Category | Tools | What it does |
|
||||||
|
|----------|-------|--------------|
|
||||||
|
| **Functions** | 11 | Decompile, disassemble, rename, set signatures, list variables |
|
||||||
|
| **Data** | 8 | Create/modify data items, list strings, set types |
|
||||||
|
| **Structs** | 7 | Create structs, add/update fields, manage data types |
|
||||||
|
| **Symbols** | 9 | Create labels, rename symbols, list imports/exports |
|
||||||
|
| **Analysis** | 6 | Call graphs, data flow, cross-references, run analysis |
|
||||||
|
| **Memory** | 2 | Read/write raw bytes |
|
||||||
|
| **Variables** | 4 | List/rename function variables, set types |
|
||||||
|
| **Bookmarks** | 3 | Create/list/delete analysis bookmarks |
|
||||||
|
| **Enums/Typedefs** | 4 | Create enum and typedef data types |
|
||||||
|
| **Namespaces** | 2 | List namespaces and classes |
|
||||||
|
| **Segments** | 1 | List memory segments with permissions |
|
||||||
|
| **Docker** | 7 | Auto-start containers, health checks, session management |
|
||||||
|
|
||||||
## Overview
|
**13 analysis prompts** for common RE workflows:
|
||||||
|
- `malware_triage` — Quick capability assessment
|
||||||
|
- `identify_crypto` — Find crypto functions and constants
|
||||||
|
- `find_authentication` — Locate auth, license checks, credentials
|
||||||
|
- `analyze_protocol` — Reverse network/file protocols
|
||||||
|
- `trace_data_flow` — Taint analysis through functions
|
||||||
|
- And 8 more specialized prompts...
|
||||||
|
|
||||||
GhydraMCP v2.1 integrates three key components:
|
**11 MCP resources** for quick enumeration without tool calls.
|
||||||
|
|
||||||
1. **Modular Ghidra Plugin**: Exposes Ghidra's powerful reverse engineering capabilities through a HATEOAS-driven REST API
|
---
|
||||||
2. **MCP Bridge**: A Python script that translates MCP requests into API calls with comprehensive type checking
|
|
||||||
3. **Multi-instance Architecture**: Connect multiple Ghidra instances to analyze different binaries simultaneously
|
|
||||||
|
|
||||||
This architecture enables AI assistants like Claude to seamlessly:
|
## Quick Start
|
||||||
- Decompile and analyze binary code with customizable output formats
|
|
||||||
- Map program structures, function relationships, and complex data types
|
|
||||||
- Perform advanced binary analysis (cross-references, call graphs, data flow, etc.)
|
|
||||||
- Make precise modifications to the analysis (rename, annotate, create/delete/modify data, etc.)
|
|
||||||
- Read memory directly and manipulate binary at a low level
|
|
||||||
- Navigate resources through discoverable HATEOAS links
|
|
||||||
|
|
||||||
GhydraMCP is based on [GhidraMCP by Laurie Wired](https://github.com/LaurieWired/GhidraMCP/) but has evolved into a comprehensive reverse engineering platform with enhanced multi-instance support, extensive data manipulation capabilities, and a robust HATEOAS-compliant API architecture.
|
### Option 1: Docker (Easiest)
|
||||||
|
|
||||||
# Features
|
No Ghidra installation needed. Analyze binaries in isolated containers.
|
||||||
|
|
||||||
GhydraMCP version 2.1 provides a comprehensive set of reverse engineering capabilities to AI assistants through its HATEOAS-driven API:
|
```bash
|
||||||
|
# Build the image (once)
|
||||||
|
cd GhydraMCP && docker build -t ghydramcp:latest -f docker/Dockerfile .
|
||||||
|
|
||||||
## Advanced Program Analysis
|
# Add to your MCP config
|
||||||
|
claude mcp add ghydramcp -- uv run --directory /path/to/GhydraMCP ghydramcp
|
||||||
|
```
|
||||||
|
|
||||||
- **Enhanced Decompilation**:
|
Then in Claude:
|
||||||
- Convert binary functions to readable C code
|
```
|
||||||
- Toggle between clean C-like pseudocode and raw decompiler output
|
Analyze /path/to/suspicious.exe
|
||||||
- Show/hide syntax trees for detailed analysis
|
```
|
||||||
- Multiple simplification styles for different analysis approaches
|
|
||||||
|
|
||||||
- **Comprehensive Static Analysis**:
|
Claude will auto-start a container, wait for analysis, and begin work.
|
||||||
- Cross-reference analysis (find callers and callees)
|
|
||||||
- Complete call graph generation and traversal
|
|
||||||
- Data flow analysis with variable tracking
|
|
||||||
- Type propagation and reconstruction
|
|
||||||
- Function relationship mapping
|
|
||||||
|
|
||||||
- **Memory Operations**:
|
### Option 2: Native Ghidra
|
||||||
- Direct memory reading with hex and raw byte representation
|
|
||||||
- Address space navigation and mapping
|
|
||||||
- Memory segment analysis
|
|
||||||
|
|
||||||
- **Symbol Management**:
|
1. **Install the Ghidra plugin:**
|
||||||
- View and analyze imports and exports
|
- Download latest [release](https://github.com/starsong-consulting/GhydraMCP/releases)
|
||||||
- Identify library functions and dependencies
|
- In Ghidra: `File → Install Extensions → +` → select the `.zip`
|
||||||
- Symbol table exploration and manipulation
|
- Restart Ghidra
|
||||||
- Namespace hierarchy visualization
|
- Enable in `File → Configure → Developer → GhydraMCPPlugin`
|
||||||
|
|
||||||
## Interactive Reverse Engineering
|
2. **Add MCP server:**
|
||||||
|
```bash
|
||||||
|
claude mcp add ghydramcp -- uv run --directory /path/to/GhydraMCP ghydramcp
|
||||||
|
```
|
||||||
|
|
||||||
- **Code Understanding**:
|
3. **Open a binary in Ghidra**, then ask Claude to analyze it.
|
||||||
- Explore function code with rich context
|
|
||||||
- Analyze data structures and complex types
|
|
||||||
- View disassembly with linking to decompiled code
|
|
||||||
- Examine function prototypes and signatures
|
|
||||||
|
|
||||||
- **Comprehensive Annotation**:
|
---
|
||||||
- Rename functions, variables, and data
|
|
||||||
- Add multiple comment types (EOL, plate, pre/post)
|
|
||||||
- Create and modify data types
|
|
||||||
- Set and update function signatures and prototypes
|
|
||||||
|
|
||||||
## Complete Data Manipulation
|
## How It Works
|
||||||
|
|
||||||
- **Data Creation and Management**:
|
```
|
||||||
- Create new data items with specified types
|
┌──────────────┐ MCP ┌──────────────┐ HTTP ┌──────────────┐
|
||||||
- Delete existing data items
|
│ Claude │◄────────────►│ GhydraMCP │◄────────────►│ Ghidra │
|
||||||
- Rename data items with proper scope handling
|
│ (or other │ stdio │ (Python) │ REST API │ Plugin │
|
||||||
- Set and update data types for existing items
|
│ MCP client) │ │ │ │ (Java) │
|
||||||
- Combined rename and retype operations
|
└──────────────┘ └──────────────┘ └──────────────┘
|
||||||
- Type definition management
|
```
|
||||||
|
|
||||||
- **Function Manipulation**:
|
- **Ghidra Plugin**: Exposes Ghidra's analysis via HTTP REST API (HATEOAS)
|
||||||
- Rename functions with proper scoping
|
- **GhydraMCP Server**: Translates MCP tool calls to API requests
|
||||||
- Update function signatures with parameter information
|
- **Multi-instance**: Analyze multiple binaries simultaneously on different ports
|
||||||
- Modify local variable names and types
|
- **Session isolation**: Docker containers get unique ports, preventing conflicts
|
||||||
- Set function return types
|
|
||||||
|
|
||||||
## Multi-instance Support
|
---
|
||||||
|
|
||||||
- Run multiple Ghidra instances simultaneously
|
## Usage Patterns
|
||||||
- Analyze different binaries in parallel
|
|
||||||
- Connect to specific instances using port numbers
|
|
||||||
- Auto-discovery of running Ghidra instances
|
|
||||||
- Instance metadata with project and file information
|
|
||||||
- Plugin version and API checking for compatibility
|
|
||||||
|
|
||||||
## Program Navigation and Discovery
|
### Set Current Instance (Then Forget About Ports)
|
||||||
|
|
||||||
- List and search functions, classes, and namespaces
|
|
||||||
- View memory segments and layout
|
|
||||||
- Search by name, pattern, or signature
|
|
||||||
- Resource discovery through HATEOAS links
|
|
||||||
- Pagination for handling large result sets
|
|
||||||
- Filtering capabilities across all resources
|
|
||||||
|
|
||||||
# Installation
|
|
||||||
|
|
||||||
## Prerequisites
|
|
||||||
- Install [Ghidra](https://ghidra-sre.org)
|
|
||||||
- Python3
|
|
||||||
- MCP [SDK](https://github.com/modelcontextprotocol/python-sdk)
|
|
||||||
|
|
||||||
## Ghidra
|
|
||||||
First, download the latest [release](https://github.com/teal-bauer/GhydraMCP/releases) from this repository. The "Complete" artifact contains the zipped Ghidra plugin and the Python MCP bridge. Unpack the outer archive, then, add the plugin to Ghidra:
|
|
||||||
|
|
||||||
1. Run Ghidra
|
|
||||||
2. Select `File` -> `Install Extensions`
|
|
||||||
3. Click the `+` button
|
|
||||||
4. Select the `GhydraMCP-[version].zip` file from the downloaded release
|
|
||||||
5. Restart Ghidra
|
|
||||||
6. Make sure the GhydraMCPPlugin is enabled in `File` -> `Configure` -> `Developer`
|
|
||||||
|
|
||||||
> **Note:** By default, the first CodeBrowser opened in Ghidra gets port 8192, the second gets 8193, and so on. You can check which ports are being used by looking at the Console in the Ghidra main (project) window - click the computer icon in the bottom right to "Open Console". Look for log entries like:
|
|
||||||
> ```
|
|
||||||
> (HydraMCPPlugin) Plugin loaded on port 8193
|
|
||||||
> (HydraMCPPlugin) HydraMCP HTTP server started on port 8193
|
|
||||||
> ```
|
|
||||||
>
|
|
||||||
> GhydraMCP now includes auto-discovery of running Ghidra instances, so manually registering each instance is typically not necessary. The MCP bridge will automatically discover and register instances on startup and periodically check for new ones.
|
|
||||||
|
|
||||||
Video Installation Guide:
|
|
||||||
|
|
||||||
https://github.com/user-attachments/assets/75f0c176-6da1-48dc-ad96-c182eb4648c3
|
|
||||||
|
|
||||||
## MCP Clients
|
|
||||||
|
|
||||||
GhydraMCP works with any MCP-compatible client using **stdio transport**. It has been tested and confirmed working with:
|
|
||||||
|
|
||||||
- **Claude Desktop** - Anthropic's official desktop application
|
|
||||||
- **Claude Code** - Anthropic's VS Code extension and CLI tool
|
|
||||||
- **Cline** - Popular VS Code extension for AI-assisted coding
|
|
||||||
|
|
||||||
See the [Client Setup](#client-setup) section below for detailed configuration instructions for each client.
|
|
||||||
|
|
||||||
## API Reference (Updated for v2.1)
|
|
||||||
|
|
||||||
### Available Tools
|
|
||||||
|
|
||||||
GhydraMCP v2.1 organizes tools into logical namespaces for better discoverability and organization:
|
|
||||||
|
|
||||||
**Instance Management** (`instances_*`):
|
|
||||||
- `instances_list`: List active Ghidra instances (auto-discovers on default host) - **use this first**
|
|
||||||
- `instances_discover`: Discover instances on a specific host (params: host [optional]) - **only use for non-default hosts**
|
|
||||||
- `instances_register`: Register new instance (params: port, url [optional])
|
|
||||||
- `instances_unregister`: Remove instance (params: port)
|
|
||||||
- `instances_use`: Set current working instance (params: port)
|
|
||||||
- `instances_current`: Get current working instance info
|
|
||||||
|
|
||||||
**Function Analysis** (`functions_*`):
|
|
||||||
- `functions_list`: List all functions (params: offset, limit, port [optional])
|
|
||||||
- `functions_get`: Get function details (params: name or address, port [optional])
|
|
||||||
- `functions_decompile`: Get decompiled C code (params: name or address, syntax_tree, style, timeout, port [optional])
|
|
||||||
- `functions_disassemble`: Get disassembled instructions (params: name or address, port [optional])
|
|
||||||
- `functions_create`: Create function at address (params: address, port [optional])
|
|
||||||
- `functions_rename`: Rename a function (params: old_name or address, new_name, port [optional])
|
|
||||||
- `functions_set_signature`: Update function prototype (params: name or address, signature, port [optional])
|
|
||||||
- `functions_get_variables`: Get function variables (params: name or address, port [optional])
|
|
||||||
- `functions_set_comment`: Set function comment (params: address, comment, port [optional])
|
|
||||||
|
|
||||||
**Data Manipulation** (`data_*`):
|
|
||||||
- `data_list`: List data items (params: offset, limit, addr, name, name_contains, port [optional])
|
|
||||||
- `data_list_strings`: List all defined strings (params: offset, limit, filter, port [optional])
|
|
||||||
- `data_create`: Create data at address (params: address, data_type, size [optional], port [optional])
|
|
||||||
- `data_rename`: Rename data item (params: address, name, port [optional])
|
|
||||||
- `data_delete`: Delete data item (params: address, port [optional])
|
|
||||||
- `data_set_type`: Change data type (params: address, data_type, port [optional])
|
|
||||||
|
|
||||||
**Struct Management** (`structs_*`):
|
|
||||||
- `structs_list`: List all struct data types (params: offset, limit, category [optional], port [optional])
|
|
||||||
- `structs_get`: Get detailed struct information (params: name, port [optional])
|
|
||||||
- `structs_create`: Create new struct (params: name, category [optional], description [optional], port [optional])
|
|
||||||
- `structs_add_field`: Add field to struct (params: struct_name, field_name, field_type, offset [optional], comment [optional], port [optional])
|
|
||||||
- `structs_update_field`: Update struct field (params: struct_name, field_name or field_offset, new_name [optional], new_type [optional], new_comment [optional], port [optional])
|
|
||||||
- `structs_delete`: Delete struct (params: name, port [optional])
|
|
||||||
|
|
||||||
**Memory Operations** (`memory_*`):
|
|
||||||
- `memory_read`: Read bytes from memory (params: address, length, format, port [optional])
|
|
||||||
- `memory_write`: Write bytes to memory (params: address, bytes_data, format, port [optional])
|
|
||||||
|
|
||||||
**Cross-References** (`xrefs_*`):
|
|
||||||
- `xrefs_list`: List cross-references (params: to_addr [optional], from_addr [optional], type [optional], offset, limit, port [optional])
|
|
||||||
|
|
||||||
**Analysis** (`analysis_*`):
|
|
||||||
- `analysis_run`: Trigger program analysis (params: port [optional], analysis_options [optional])
|
|
||||||
- `analysis_get_callgraph`: Get function call graph (params: name or address, max_depth, port [optional])
|
|
||||||
- `analysis_get_dataflow`: Perform data flow analysis (params: address, direction, max_steps, port [optional])
|
|
||||||
|
|
||||||
**Example Usage**:
|
|
||||||
```python
|
```python
|
||||||
# Instance Management - Always start here
|
instances_list() # Discover running Ghidra instances
|
||||||
client.use_tool("ghydra", "instances_list") # Auto-discovers instances on localhost
|
instances_use(port=8192) # Set as current
|
||||||
client.use_tool("ghydra", "instances_use", {"port": 8192}) # Set working instance
|
functions_list() # No port needed!
|
||||||
client.use_tool("ghydra", "instances_current") # Check current instance
|
data_list_strings(grep="password") # Uses current instance
|
||||||
|
|
||||||
# Function Analysis
|
|
||||||
client.use_tool("ghydra", "functions_list", {"offset": 0, "limit": 100})
|
|
||||||
client.use_tool("ghydra", "functions_get", {"name": "main"})
|
|
||||||
client.use_tool("ghydra", "functions_decompile", {"address": "0x00401000"})
|
|
||||||
client.use_tool("ghydra", "functions_disassemble", {"name": "main"})
|
|
||||||
client.use_tool("ghydra", "functions_rename", {"address": "0x00401000", "new_name": "process_data"})
|
|
||||||
client.use_tool("ghydra", "functions_set_signature", {"address": "0x00401000", "signature": "int process_data(char* buf, int len)"})
|
|
||||||
client.use_tool("ghydra", "functions_set_comment", {"address": "0x00401000", "comment": "Main processing function"})
|
|
||||||
|
|
||||||
# Data Manipulation
|
|
||||||
client.use_tool("ghydra", "data_list_strings", {"filter": "password"}) # Find strings containing "password"
|
|
||||||
client.use_tool("ghydra", "data_list", {"offset": 0, "limit": 50})
|
|
||||||
client.use_tool("ghydra", "data_create", {"address": "0x00401234", "data_type": "int"})
|
|
||||||
client.use_tool("ghydra", "data_rename", {"address": "0x00401234", "name": "counter"})
|
|
||||||
client.use_tool("ghydra", "data_set_type", {"address": "0x00401238", "data_type": "char *"})
|
|
||||||
client.use_tool("ghydra", "data_delete", {"address": "0x0040123C"})
|
|
||||||
|
|
||||||
# Struct Management
|
|
||||||
client.use_tool("ghydra", "structs_create", {"name": "NetworkPacket", "category": "/network"})
|
|
||||||
client.use_tool("ghydra", "structs_add_field", {
|
|
||||||
"struct_name": "NetworkPacket",
|
|
||||||
"field_name": "header",
|
|
||||||
"field_type": "dword",
|
|
||||||
"comment": "Packet header"
|
|
||||||
})
|
|
||||||
client.use_tool("ghydra", "structs_add_field", {
|
|
||||||
"struct_name": "NetworkPacket",
|
|
||||||
"field_name": "data_ptr",
|
|
||||||
"field_type": "pointer"
|
|
||||||
})
|
|
||||||
client.use_tool("ghydra", "structs_update_field", {
|
|
||||||
"struct_name": "NetworkPacket",
|
|
||||||
"field_name": "header",
|
|
||||||
"new_name": "packet_header",
|
|
||||||
"new_comment": "Updated header field"
|
|
||||||
})
|
|
||||||
client.use_tool("ghydra", "structs_get", {"name": "NetworkPacket"})
|
|
||||||
client.use_tool("ghydra", "structs_list", {"category": "/network"})
|
|
||||||
|
|
||||||
# Memory Operations
|
|
||||||
client.use_tool("ghydra", "memory_read", {"address": "0x00401000", "length": 16, "format": "hex"})
|
|
||||||
client.use_tool("ghydra", "memory_write", {"address": "0x00401000", "bytes_data": "90909090", "format": "hex"})
|
|
||||||
|
|
||||||
# Cross-References
|
|
||||||
client.use_tool("ghydra", "xrefs_list", {"to_addr": "0x00401000"}) # Find callers
|
|
||||||
client.use_tool("ghydra", "xrefs_list", {"from_addr": "0x00401000"}) # Find callees
|
|
||||||
|
|
||||||
# Analysis
|
|
||||||
client.use_tool("ghydra", "analysis_get_callgraph", {"name": "main", "max_depth": 5})
|
|
||||||
client.use_tool("ghydra", "analysis_get_dataflow", {"address": "0x00401050", "direction": "forward"})
|
|
||||||
client.use_tool("ghydra", "analysis_run") # Trigger full analysis
|
|
||||||
```
|
```
|
||||||
|
|
||||||
## Client Setup
|
### Docker Workflow
|
||||||
|
|
||||||
GhydraMCP works with any MCP-compatible client. Below are configuration examples for popular AI coding assistants.
|
```python
|
||||||
|
# Start container (returns immediately)
|
||||||
|
result = docker_auto_start(binary_path="/path/to/malware.exe")
|
||||||
|
# → {port: 8195, message: "Poll docker_health(port=8195)..."}
|
||||||
|
|
||||||
### Installation Methods
|
# Poll until ready
|
||||||
|
while True:
|
||||||
|
health = docker_health(port=8195)
|
||||||
|
if health["healthy"]:
|
||||||
|
break
|
||||||
|
# Can check docker_logs() while waiting
|
||||||
|
|
||||||
#### Recommended: Local Installation from Release
|
# Register and use
|
||||||
|
instances_use(port=8195)
|
||||||
|
functions_list() # Ready to analyze
|
||||||
|
```
|
||||||
|
|
||||||
Download the latest [release](https://github.com/starsong-consulting/GhydraMCP/releases) to ensure the bridge and plugin versions are in sync.
|
### Cursor-Based Pagination
|
||||||
|
|
||||||
|
Large binaries can have 100K+ functions. Use cursors:
|
||||||
|
|
||||||
|
```python
|
||||||
|
result = functions_list(page_size=100)
|
||||||
|
# → {items: [...], cursor_id: "abc123", has_more: true}
|
||||||
|
|
||||||
|
# Get next page
|
||||||
|
cursor_next(cursor_id="abc123")
|
||||||
|
|
||||||
|
# Or filter server-side
|
||||||
|
functions_list(grep="crypto|encrypt", page_size=50)
|
||||||
|
```
|
||||||
|
|
||||||
|
### Analysis Prompts
|
||||||
|
|
||||||
|
Built-in prompts for common workflows:
|
||||||
|
|
||||||
|
```
|
||||||
|
/prompt malware_triage
|
||||||
|
/prompt identify_crypto
|
||||||
|
/prompt find_authentication
|
||||||
|
```
|
||||||
|
|
||||||
|
These guide Claude through systematic analysis with progress reporting.
|
||||||
|
|
||||||
|
---
|
||||||
|
|
||||||
|
## Configuration
|
||||||
|
|
||||||
|
### Environment Variables
|
||||||
|
|
||||||
|
| Variable | Default | Description |
|
||||||
|
|----------|---------|-------------|
|
||||||
|
| `GHIDRA_HYDRA_HOST` | `localhost` | Ghidra instance host |
|
||||||
|
| `GHIDRA_HYDRA_PORT` | `8192` | Default port |
|
||||||
|
|
||||||
|
### MCP Config Examples
|
||||||
|
|
||||||
|
**Claude Desktop** (`~/Library/Application Support/Claude/claude_desktop_config.json`):
|
||||||
```json
|
```json
|
||||||
{
|
{
|
||||||
"mcpServers": {
|
"mcpServers": {
|
||||||
"ghydra": {
|
"ghydramcp": {
|
||||||
"command": "uv",
|
"command": "uv",
|
||||||
"args": [
|
"args": ["run", "--directory", "/path/to/GhydraMCP", "ghydramcp"]
|
||||||
"run",
|
|
||||||
"/ABSOLUTE_PATH_TO/bridge_mcp_hydra.py"
|
|
||||||
],
|
|
||||||
"env": {
|
|
||||||
"GHIDRA_HYDRA_HOST": "localhost"
|
|
||||||
}
|
|
||||||
}
|
|
||||||
}
|
|
||||||
}
|
|
||||||
```
|
|
||||||
|
|
||||||
Replace `/ABSOLUTE_PATH_TO/` with the actual path to your `bridge_mcp_hydra.py` file.
|
|
||||||
|
|
||||||
> **Note:** You can also use `python` instead of `uv run`, but then you'll need to manually install the requirements first with `pip install mcp requests`.
|
|
||||||
|
|
||||||
#### Alternative: Direct from Repository with uvx
|
|
||||||
|
|
||||||
If you want to use the latest development version, you can run directly from the GitHub repository:
|
|
||||||
|
|
||||||
```json
|
|
||||||
{
|
|
||||||
"mcpServers": {
|
|
||||||
"ghydra": {
|
|
||||||
"command": "uvx",
|
|
||||||
"args": [
|
|
||||||
"--from",
|
|
||||||
"git+https://github.com/starsong-consulting/GhydraMCP",
|
|
||||||
"ghydramcp"
|
|
||||||
],
|
|
||||||
"env": {
|
|
||||||
"GHIDRA_HYDRA_HOST": "localhost"
|
|
||||||
}
|
|
||||||
}
|
|
||||||
}
|
|
||||||
}
|
|
||||||
```
|
|
||||||
|
|
||||||
> **Warning:** This method may pull a bridge version that's out of sync with your installed plugin. Only use this if you're tracking the latest development branch.
|
|
||||||
|
|
||||||
### Claude Desktop Configuration
|
|
||||||
|
|
||||||
Add your chosen configuration method to your Claude Desktop configuration file:
|
|
||||||
- **macOS**: `~/Library/Application Support/Claude/claude_desktop_config.json`
|
|
||||||
- **Windows**: `%APPDATA%\Claude\claude_desktop_config.json`
|
|
||||||
|
|
||||||
### Claude Code Configuration
|
|
||||||
|
|
||||||
Claude Code automatically discovers MCP servers configured in Claude Desktop. If you've set up the configuration above, Claude Code will have access to GhydraMCP tools immediately.
|
|
||||||
|
|
||||||
Alternatively, you can configure Claude Code separately by adding the same configuration to the MCP settings in Claude Code's configuration.
|
|
||||||
|
|
||||||
### Cline Configuration
|
|
||||||
|
|
||||||
Cline (VS Code extension) uses a separate configuration file. To set up GhydraMCP with Cline:
|
|
||||||
|
|
||||||
1. Open VS Code with Cline installed
|
|
||||||
2. Click the "MCP Servers" icon in Cline's interface
|
|
||||||
3. Select the "Configure" tab
|
|
||||||
4. Click "Configure MCP Servers" to edit `cline_mcp_settings.json`
|
|
||||||
5. Add the following configuration:
|
|
||||||
|
|
||||||
```json
|
|
||||||
{
|
|
||||||
"mcpServers": {
|
|
||||||
"ghydra": {
|
|
||||||
"command": "uv",
|
|
||||||
"args": [
|
|
||||||
"run",
|
|
||||||
"/ABSOLUTE_PATH_TO/bridge_mcp_hydra.py"
|
|
||||||
],
|
|
||||||
"env": {
|
|
||||||
"GHIDRA_HYDRA_HOST": "localhost"
|
|
||||||
},
|
|
||||||
"disabled": false
|
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
```
|
```
|
||||||
|
|
||||||
If you prefer to use `python` directly instead of `uv`:
|
**Claude Code**:
|
||||||
|
```bash
|
||||||
```json
|
claude mcp add ghydramcp -- uv run --directory /path/to/GhydraMCP ghydramcp
|
||||||
{
|
|
||||||
"mcpServers": {
|
|
||||||
"ghydra": {
|
|
||||||
"command": "python",
|
|
||||||
"args": [
|
|
||||||
"/ABSOLUTE_PATH_TO/bridge_mcp_hydra.py"
|
|
||||||
],
|
|
||||||
"env": {
|
|
||||||
"GHIDRA_HYDRA_HOST": "localhost"
|
|
||||||
},
|
|
||||||
"disabled": false
|
|
||||||
}
|
|
||||||
}
|
|
||||||
}
|
|
||||||
```
|
```
|
||||||
|
|
||||||
> **Important:** Replace `/ABSOLUTE_PATH_TO/` with the actual absolute path to your `bridge_mcp_hydra.py` file. For example:
|
---
|
||||||
> - **Linux/macOS**: `/home/username/GhydraMCP/bridge_mcp_hydra.py`
|
|
||||||
> - **Windows**: `C:\\Users\\username\\GhydraMCP\\bridge_mcp_hydra.py`
|
|
||||||
|
|
||||||
After saving the configuration, restart Cline to load the GhydraMCP server.
|
## Tool Reference
|
||||||
|
|
||||||
### 5ire Configuration
|
### Instance Management
|
||||||
1. Tool Key: ghydra
|
|
||||||
2. Name: GhydraMCP
|
|
||||||
3. Command: `uv run /ABSOLUTE_PATH_TO/bridge_mcp_hydra.py`
|
|
||||||
|
|
||||||
## Example Session
|
|
||||||
|
|
||||||
Below is an example of how an AI assistant might interact with Ghidra using GhydraMCP v2.1:
|
|
||||||
|
|
||||||
**User:**
|
|
||||||
```
|
```
|
||||||
Analyze the main function in the binary and identify any data structures being used.
|
instances_list # Discover Ghidra instances (use this first!)
|
||||||
|
instances_use # Set current working instance
|
||||||
|
instances_current # Show current instance info
|
||||||
```
|
```
|
||||||
|
|
||||||
**Assistant:**
|
### Function Analysis
|
||||||
```
|
```
|
||||||
I'll analyze the binary for you. Let me start by checking available Ghidra instances.
|
functions_list # List functions (supports grep, pagination)
|
||||||
|
functions_get # Get function details by name or address
|
||||||
[Uses instances_list]
|
functions_decompile # Decompile to C pseudocode
|
||||||
Found Ghidra instance on port 8192 analyzing "malware_sample.exe".
|
functions_disassemble # Get assembly instructions
|
||||||
|
functions_rename # Rename a function
|
||||||
[Uses instances_use with port 8192]
|
functions_set_signature # Set function prototype
|
||||||
Set instance 8192 as current working instance.
|
functions_set_comment # Add decompiler comment
|
||||||
|
functions_create # Create function at address
|
||||||
[Uses functions_decompile with name "main"]
|
functions_variables # List local variables and parameters
|
||||||
Here's the decompiled main function:
|
|
||||||
|
|
||||||
void main(void) {
|
|
||||||
NetworkPacket *packet;
|
|
||||||
int status;
|
|
||||||
|
|
||||||
packet = (NetworkPacket *)malloc(0x20);
|
|
||||||
packet->header = 0x1234;
|
|
||||||
packet->data_ptr = allocate_buffer(0x100);
|
|
||||||
status = process_packet(packet);
|
|
||||||
...
|
|
||||||
}
|
|
||||||
|
|
||||||
I can see this code uses a "NetworkPacket" structure. Let me check if it's already defined.
|
|
||||||
|
|
||||||
[Uses structs_get with name "NetworkPacket"]
|
|
||||||
The struct doesn't exist yet. Let me create it based on the decompiled code.
|
|
||||||
|
|
||||||
[Uses structs_create with name "NetworkPacket", category "/network"]
|
|
||||||
Created NetworkPacket struct.
|
|
||||||
|
|
||||||
[Uses structs_add_field for each field]
|
|
||||||
Added fields:
|
|
||||||
- header (dword at offset 0)
|
|
||||||
- data_ptr (pointer at offset 4)
|
|
||||||
- size (dword at offset 8)
|
|
||||||
- status (dword at offset 12)
|
|
||||||
|
|
||||||
[Uses data_list_strings with filter "Error"]
|
|
||||||
I also found several error strings that suggest this binary handles network communications with error checking.
|
|
||||||
|
|
||||||
The main function appears to be the entry point for a network packet processing routine that allocates and processes data structures for network communication.
|
|
||||||
```
|
```
|
||||||
|
|
||||||
**User:**
|
### Data Operations
|
||||||
```
|
```
|
||||||
Can you trace where process_packet is called from and show me the call graph?
|
data_list # List defined data items
|
||||||
|
data_list_strings # List strings (with grep filtering)
|
||||||
|
data_create # Define data at address
|
||||||
|
data_rename # Rename data item
|
||||||
|
data_set_type # Change data type
|
||||||
|
data_delete # Remove data definition
|
||||||
```
|
```
|
||||||
|
|
||||||
**Assistant:**
|
### Cross-References & Analysis
|
||||||
```
|
```
|
||||||
I'll trace the call relationships for process_packet.
|
xrefs_list # Find cross-references to/from address
|
||||||
|
analysis_get_callgraph # Generate call graph
|
||||||
[Uses xrefs_list with to_addr for process_packet]
|
analysis_get_dataflow # Trace data flow forward/backward
|
||||||
Found 3 callers of process_packet:
|
analysis_run # Trigger Ghidra auto-analysis
|
||||||
- main at 0x401000
|
|
||||||
- worker_thread at 0x401200
|
|
||||||
- handle_request at 0x401450
|
|
||||||
|
|
||||||
[Uses analysis_get_callgraph with name "process_packet", max_depth 3]
|
|
||||||
Here's the call graph starting from process_packet:
|
|
||||||
|
|
||||||
process_packet (0x401100)
|
|
||||||
├── validate_header (0x401150)
|
|
||||||
│ └── check_magic (0x401180)
|
|
||||||
├── parse_data (0x4011A0)
|
|
||||||
│ ├── extract_field (0x4011D0)
|
|
||||||
│ └── validate_checksum (0x4011E0)
|
|
||||||
└── send_response (0x401220)
|
|
||||||
└── network_send (0x401250)
|
|
||||||
|
|
||||||
This shows process_packet coordinates validation, parsing, and response transmission.
|
|
||||||
```
|
```
|
||||||
|
|
||||||
# JSON Communication
|
### Structs & Types
|
||||||
|
```
|
||||||
GhydraMCP uses structured JSON for all communication between the Python bridge and Java plugin. This ensures consistent and reliable data exchange.
|
structs_list # List struct definitions
|
||||||
|
structs_get # Get struct with all fields
|
||||||
## API Architecture
|
structs_create # Create new struct
|
||||||
|
structs_add_field # Add field to struct
|
||||||
GhydraMCP v2.1 implements a comprehensive HATEOAS-driven REST API that follows hypermedia design principles:
|
structs_update_field # Modify existing field
|
||||||
|
structs_delete # Remove struct
|
||||||
### Core API Design
|
enums_list / enums_create
|
||||||
|
typedefs_list / typedefs_create
|
||||||
- **HATEOAS Architecture**: Each response includes navigational links for resource discovery
|
|
||||||
- **Versioned Endpoints**: All requests verified against API version for compatibility
|
|
||||||
- **Structured Responses**: Standardized JSON format with consistent field naming
|
|
||||||
- **Proper HTTP Methods**: GET for retrieval, POST for creation, PATCH for updates, DELETE for removal
|
|
||||||
- **Appropriate Status Codes**: Uses standard HTTP status codes for clear error handling
|
|
||||||
|
|
||||||
### Response Format
|
|
||||||
|
|
||||||
All responses follow this HATEOAS-driven format:
|
|
||||||
|
|
||||||
```json
|
|
||||||
{
|
|
||||||
"id": "req-123",
|
|
||||||
"instance": "http://localhost:8192",
|
|
||||||
"success": true,
|
|
||||||
"result": "...",
|
|
||||||
"timestamp": 1712159482123,
|
|
||||||
"_links": {
|
|
||||||
"self": {"href": "/endpoint/current"},
|
|
||||||
"related": [
|
|
||||||
{"href": "/endpoint/related1", "name": "Related Resource 1"},
|
|
||||||
{"href": "/endpoint/related2", "name": "Related Resource 2"}
|
|
||||||
]
|
|
||||||
}
|
|
||||||
}
|
|
||||||
```
|
```
|
||||||
|
|
||||||
For list responses, pagination information is included:
|
### Docker Management
|
||||||
|
```
|
||||||
```json
|
docker_auto_start # Start container for binary (auto port allocation)
|
||||||
{
|
docker_health # Check if container API is responding
|
||||||
"id": "req-123",
|
docker_status # List all containers and images
|
||||||
"instance": "http://localhost:8192",
|
docker_start # Manual container start
|
||||||
"success": true,
|
docker_stop # Stop container (session-scoped)
|
||||||
"result": [ ... objects ... ],
|
docker_logs # Get container logs
|
||||||
"size": 150,
|
docker_cleanup # Remove orphaned containers
|
||||||
"offset": 0,
|
|
||||||
"limit": 50,
|
|
||||||
"_links": {
|
|
||||||
"self": { "href": "/functions?offset=0&limit=50" },
|
|
||||||
"next": { "href": "/functions?offset=50&limit=50" },
|
|
||||||
"prev": { "href": "/functions?offset=0&limit=50" }
|
|
||||||
}
|
|
||||||
}
|
|
||||||
```
|
```
|
||||||
|
|
||||||
Error responses include detailed information:
|
See `--help` or the [API docs](GHIDRA_HTTP_API.md) for full parameter details.
|
||||||
|
|
||||||
```json
|
---
|
||||||
{
|
|
||||||
"id": "req-123",
|
|
||||||
"instance": "http://localhost:8192",
|
|
||||||
"success": false,
|
|
||||||
"error": {
|
|
||||||
"code": "RESOURCE_NOT_FOUND",
|
|
||||||
"message": "Function 'main' not found in current program"
|
|
||||||
},
|
|
||||||
"status_code": 404,
|
|
||||||
"timestamp": 1712159482123,
|
|
||||||
"_links": {
|
|
||||||
"self": {"href": "/functions/main"}
|
|
||||||
}
|
|
||||||
}
|
|
||||||
```
|
|
||||||
|
|
||||||
This HATEOAS approach enables resource discovery and self-documenting APIs, making integration and exploration significantly easier.
|
## Building from Source
|
||||||
|
|
||||||
# Testing
|
```bash
|
||||||
|
# Clone
|
||||||
|
git clone https://github.com/starsong-consulting/GhydraMCP
|
||||||
|
cd GhydraMCP
|
||||||
|
|
||||||
GhydraMCP includes comprehensive test suites for both the HTTP API and MCP bridge. See [TESTING.md](TESTING.md) for details on running the tests.
|
# Build Ghidra plugin
|
||||||
|
|
||||||
## HTTP API Tests
|
|
||||||
|
|
||||||
Tests the HTTP endpoints exposed by the Java plugin:
|
|
||||||
- Response format and structure
|
|
||||||
- JSON structure consistency
|
|
||||||
- Required fields in responses
|
|
||||||
- Error handling
|
|
||||||
|
|
||||||
## MCP Bridge Tests
|
|
||||||
|
|
||||||
Tests the MCP bridge functionality:
|
|
||||||
- MCP protocol communication
|
|
||||||
- Tool availability and structure
|
|
||||||
- Response format and structure
|
|
||||||
- JSON structure consistency
|
|
||||||
|
|
||||||
# Building from Source
|
|
||||||
|
|
||||||
You can build different artifacts with Maven:
|
|
||||||
|
|
||||||
## Build Everything (Default)
|
|
||||||
Build both the Ghidra plugin and the complete package:
|
|
||||||
|
|
||||||
```
|
|
||||||
mvn clean package
|
mvn clean package
|
||||||
|
# → target/GhydraMCP-[version].zip
|
||||||
|
|
||||||
|
# Build Docker image
|
||||||
|
docker build -t ghydramcp:latest -f docker/Dockerfile .
|
||||||
|
|
||||||
|
# Run MCP server (for development)
|
||||||
|
uv run ghydramcp
|
||||||
```
|
```
|
||||||
|
|
||||||
This creates:
|
---
|
||||||
- `target/GhydraMCP-[version].zip` - The Ghidra plugin only
|
|
||||||
- `target/GhydraMCP-Complete-[version].zip` - Complete package with plugin and bridge script
|
|
||||||
|
|
||||||
## Build Ghidra Plugin Only
|
## Architecture
|
||||||
If you only need the Ghidra plugin:
|
|
||||||
|
|
||||||
```
|
GhydraMCP is designed for AI agents:
|
||||||
mvn clean package -P plugin-only
|
|
||||||
```
|
|
||||||
|
|
||||||
## Build Complete Package Only
|
- **Lazy registration**: `instances_use` doesn't block — validates on first real call
|
||||||
If you only need the combined package:
|
- **Non-blocking I/O**: All Docker/HTTP operations run in thread executors
|
||||||
|
- **Session isolation**: Each MCP session gets unique container ports
|
||||||
|
- **Cursor pagination**: Handle 100K+ item responses without context overflow
|
||||||
|
- **Server-side grep**: Filter results before they hit the wire
|
||||||
|
|
||||||
```
|
Based on [GhidraMCP by Laurie Wired](https://github.com/LaurieWired/GhidraMCP/), evolved into a comprehensive RE platform.
|
||||||
mvn clean package -P complete-only
|
|
||||||
```
|
|
||||||
|
|
||||||
The Ghidra plugin includes these files required for Ghidra to recognize the extension:
|
---
|
||||||
- lib/GhydraMCP.jar
|
|
||||||
- extension.properties
|
## License
|
||||||
- Module.manifest
|
|
||||||
|
Apache 2.0
|
||||||
|
|||||||
Loading…
x
Reference in New Issue
Block a user