Imported from reason-machines/mcp-skills (
skills/ida-no-mcp-decompiler-exporter/SKILL.md). Install upstream withnpx skills add reason-machines/mcp-skills --skill ida-no-mcp-decompiler-exporter. Copyright stays with the author.
IDA NO MCP Decompiler Exporter
Skill by ara.so — MCP Skills collection.
IDA-NO-MCP is a plugin for IDA Pro that exports decompiled code, disassembly, strings, imports, exports, and memory dumps into AI-friendly formats. Instead of complex MCP integrations, it generates organized source files that can be directly analyzed by AI IDEs like Cursor or Claude Code.
What It Does
- Exports decompiled C code: Each function as a separate
.cfile with metadata (address, callers, callees) - Automatic fallback: Falls back to disassembly (
.asm) when decompilation fails - Memory dumps: Exports all memory segments as hexdump files (1MB chunks)
- Metadata extraction: Strings, imports, exports tables
- Smart filtering: Skips library functions and invalid functions automatically
- Detailed logging: Tracks successes, fallbacks, failures, and skipped functions
Installation
Plugin Mode (Recommended)
-
Copy
INP.pyto your IDA plugins directory:- Windows:
%APPDATA%\Hex-Rays\IDA Pro\plugins\ - Linux/macOS:
~/.idapro/plugins/
- Windows:
-
Restart IDA Pro
-
Use the plugin:
- Shortcut:
Ctrl-Shift-E(quick export) - Menu:
Edit→Plugins→Export for AI
- Shortcut:
Script Mode
Run INP.py directly from IDA's script window (Alt-F7) or via command line:
# From IDA script window
execfile('/path/to/INP.py')
Output Structure
After export, the IDB directory contains:
your_binary.idb/
├── decompile/ # Decompiled C code (.c files)
├── disassembly/ # Fallback assembly (.asm files)
├── memory/ # Memory dumps (hexdump format)
├── strings.txt # All strings with addresses
├── imports.txt # Import table
├── exports.txt # Export table
├── disassembly_fallback.txt # List of fallback functions
├── decompile_failed.txt # Complete failures
└── decompile_skipped.txt # Skipped library/invalid functions
Function Export Format
Each exported function includes metadata headers:
/*
* func-name: sub_401000
* func-address: 0x401000
* export-type: decompile
* callers: 0x402000, 0x403000
* callees: 0x404000, 0x405000
*/
__int64 __fastcall sub_401000(__int64 a1, int a2)
{
// Decompiled code here
return result;
}
For disassembly fallback (.asm files):
/*
* func-name: sub_401000
* func-address: 0x401000
* export-type: disassembly
* callers: 0x402000, 0x403000
* callees: 0x404000, 0x405000
*/
sub_401000 proc near
push rbp
mov rbp, rsp
; ... assembly code
ret
sub_401000 endp
Key Plugin Components
Core Export Logic
import idaapi
import idc
import idautils
import ida_hexrays
import ida_funcs
import os
def export_decompiled_code(func_ea):
"""Export decompiled code for a function"""
try:
# Get function name
func_name = idc.get_func_name(func_ea)
# Get callers and callees
callers = [hex(xref.frm) for xref in idautils.XrefsTo(func_ea, 0)]
callees = []
for item_ea in idautils.FuncItems(func_ea):
for xref in idautils.XrefsFrom(item_ea, 0):
if xref.type in [ida_xref.fl_CN, ida_xref.fl_CF]:
callees.append(hex(xref.to))
# Try decompilation
cfunc = idaapi.decompile(func_ea)
if cfunc:
decompiled = str(cfunc)
# Build metadata header
header = f"""/*
* func-name: {func_name}
* func-address: {hex(func_ea)}
* export-type: decompile
* callers: {', '.join(callers) if callers else 'none'}
* callees: {', '.join(callees) if callees else 'none'}
*/
"""
return header + decompiled
except Exception as e:
print(f"Decompilation failed for {hex(func_ea)}: {e}")
return None
Memory Export
def export_memory_segment(seg_ea, output_dir):
"""Export memory segment as hexdump"""
seg = idaapi.getseg(seg_ea)
if not seg:
return
seg_start = seg.start_ea
seg_end = seg.end_ea
seg_size = seg_end - seg_start
max_size = 1024 * 1024 # 1MB chunks
chunk_num = 0
while seg_start < seg_end:
chunk_end = min(seg_start + max_size, seg_end)
filename = f"{hex(seg_start)}--{hex(chunk_end)}.txt"
with open(os.path.join(output_dir, filename), 'w') as f:
addr = seg_start
while addr < chunk_end:
# Read 16 bytes per line
line_bytes = []
ascii_chars = []
for i in range(16):
if addr + i >= chunk_end:
break
byte = idc.get_wide_byte(addr + i)
line_bytes.append(f"{byte:02X}")
ascii_chars.append(chr(byte) if 32 <= byte <= 126 else '.')
# Format: ADDRESS | HEX BYTES | ASCII
hex_part = ' '.join(line_bytes).ljust(48)
ascii_part = ''.join(ascii_chars)
f.write(f"{hex(addr)} | {hex_part} | {ascii_part}\n")
addr += 16
seg_start = chunk_end
chunk_num += 1
String Extraction
def export_strings(output_file):
"""Export all strings with metadata"""
with open(output_file, 'w', encoding='utf-8') as f:
strings = idautils.Strings()
for s in strings:
# Format: address, length, type, content
str_type = {
0: "ASCII",
1: "UTF-16LE",
2: "UTF-32LE"
}.get(s.strtype, "UNKNOWN")
f.write(f"{hex(s.ea)} | len={s.length} | {str_type} | {str(s)}\n")
Common Usage Patterns
Analyzing Exported Code with AI
After exporting, open the IDB directory in your AI IDE:
- Context-aware analysis: AI can read all
.cfiles and understand function relationships via caller/callee metadata - Vulnerability hunting: Ask AI to find buffer overflows, use-after-free, etc.
- Crypto detection: Identify cryptographic functions and constants
- Protocol analysis: Understand network protocol parsing logic
Adding Extra Context
Create additional directories alongside exports:
your_binary.idb/
├── decompile/ # Auto-generated
├── docs/ # Your reverse engineering notes
├── codes/ # Frida scripts, exploits, tools
└── apk/ # APK decompilation (for Android)
AI tools will index all content for comprehensive analysis.
Programmatic Integration
# Run export from IDA Python script
import INP
# Trigger export programmatically
INP.main() # Runs the full export process
# Or customize export paths
output_dir = "/custom/path/output"
INP.export_all(output_dir)
Configuration
The plugin works out-of-the-box with defaults but can be customized by editing INP.py:
# Skip library functions (default: True)
SKIP_LIB_FUNCS = True
# Maximum memory chunk size in bytes
MAX_CHUNK_SIZE = 1024 * 1024 # 1MB
# Progress reporting interval
PROGRESS_INTERVAL = 100 # Report every 100 functions
# Export types to include
EXPORT_DECOMPILE = True
EXPORT_DISASM_FALLBACK = True
EXPORT_MEMORY = True
EXPORT_STRINGS = True
EXPORT_IMPORTS = True
EXPORT_EXPORTS = True
Troubleshooting
Plugin doesn't appear in menu
- Verify
INP.pyis in the correct plugins directory - Check IDA output window for Python errors
- Ensure IDA has Hex-Rays Decompiler installed (for decompilation feature)
Decompilation fails for all functions
- Check if Hex-Rays Decompiler is licensed and active
- Some architectures may not support decompilation (will auto-fallback to disassembly)
- Check
decompile_failed.txtfor specific error messages
Out of memory during export
- Large binaries may need chunked processing
- Reduce
MAX_CHUNK_SIZEin the script - Export specific function ranges instead of entire binary
Special characters in function names
The plugin automatically sanitizes filenames:
- Replaces
/\:*?"<>|with underscores - Appends address suffix for duplicate names (e.g.,
main_401000.c)
Missing callers/callees data
- Ensure IDA has completed auto-analysis (Wait for "AU: idle" in status bar)
- Run "Reanalyze program" from Edit menu if needed
- Check if functions are properly recognized (Edit → Functions → Reanalyze program)
Integration with AI Workflows
Example: Finding vulnerabilities
# After export, ask AI in your IDE:
"Analyze all functions in decompile/ for buffer overflow vulnerabilities"
"Find all memcpy/strcpy calls and check bounds validation"
Example: Understanding malware behavior
"Trace the execution flow starting from entry point at 0x401000"
"Identify anti-debugging checks and obfuscation techniques"
"Extract C2 communication URLs from strings.txt and related functions"
Example: Protocol reverse engineering
"Find packet parsing functions using imports.txt and decompiled code"
"Document the binary protocol structure based on recv/send call patterns"
Advanced Tips
- Incremental analysis: Export once, iterate with AI on specific function subsets
- Version control: Commit exports to track understanding evolution
- Cross-reference: Combine with dynamic analysis (Frida logs, traces)
- Custom scripts: Write Python scripts that parse the exported metadata for automated analysis