# IDA Pro MCP

> Let an LLM read and annotate a binary inside your IDA database

- Category: Agents & Automation / Security Agents
- Language: Python
- Licence: MIT
- Repository: https://github.com/mrexodia/ida-pro-mcp
- Website: https://plugins.hex-rays.com/mrexodia/ida-pro-mcp
- Canonical page: https://ai-tldr.dev/tools/ida-pro-mcp/

## In short

IDA Pro MCP connects IDA Pro to an LLM client over the Model Context Protocol. It sits in the Security Agents category of the Agents & Automation tools and is open-source, written in Python and licensed MIT. To get started: activate idalib for your IDA install.

## Overview

IDA Pro MCP connects IDA Pro to an LLM client over the Model Context Protocol. It installs as an IDA plugin and exposes the parts of the disassembler an analyst actually uses — decompilation, disassembly, cross-references, the call graph, basic blocks, byte and string reads — as MCP tools an agent can call while you work.

Crucially it is not read-only. The server also exposes the annotation side of reverse engineering: renaming functions and variables, setting comments, declaring and inferring types, defining functions and code, editing stack frames and reading structures. That means an agent can work through an unfamiliar binary and leave its conclusions in the IDA database, where they persist for you rather than living in a chat log.

An optional debugger extension adds runtime control — starting and exiting the debugger, continuing, stepping, and managing breakpoints — so the same agent can drive dynamic analysis alongside the static view.

## What it does

- Static analysis tools: decompile, disasm, lookup_funcs, xrefs_to, callees, callgraph and basic_blocks
- Search across the database: find_regex, find_bytes, find_insns and a general find
- Write-back annotation: rename, set_comments, set_type, infer_types, declare_type, define_func and define_code
- Memory and data access: get_bytes, get_int, get_string, get_global_value and struct reads
- Optional debugger extension for breakpoints, stepping and reading or writing live memory
- Works with any MCP client; ships a Claude Code plugin via the author's marketplace

## Getting started

IDA Pro MCP needs IDA Pro 8.3 or newer (9 recommended — IDA Free is not supported), Python 3.11+ inside IDA, the uv package manager, and idalib activated globally.

1. Activate idalib for your IDA install

   Run the activation script that ships with IDA. Adjust the path to match your version and platform.

```bash
# macOS
uv run "/Applications/IDA Professional 9.3.app/Contents/MacOS/idalib/python/py-activate-idalib.py"
# Linux
uv run "/path/to/idapro-9.3/idalib/python/py-activate-idalib.py"
# Windows
uv run "C:\Program Files\IDA Professional 9.3\idalib\python\py-activate-idalib.py"
```

2. Install the plugin and wire up your MCP client

   The --install flag registers the IDA plugin and writes the MCP configuration for the clients it detects.

```bash
pip install https://github.com/mrexodia/ida-pro-mcp/archive/refs/heads/main.zip
ida-pro-mcp --install
```

3. Or install it as a Claude Code plugin

   The author publishes a marketplace entry, which is the quickest route if Claude Code is your client.

```bash
claude plugin marketplace add mrexodia/claude-marketplace
claude plugin install ida-pro-mcp@mrexodia
```

4. Open a binary and ask

   With a database open in IDA, the MCP tools become available to your client. Ask it to decompile a function, trace callers, or rename and comment a region it has worked out — the changes land in the IDA database.

## When to use it

- Reach for it when triaging an unfamiliar binary and you want a first pass of named functions and comments before you read it yourself
- Reach for it for malware analysis, where an agent can follow cross-references and summarise what each branch does
- Reach for it when recovering types and structures, using infer_types and declare_type to rebuild a struct layout
- Reach for it in CTF reversing, where the loop of decompile, rename, re-read is exactly what an agent can iterate quickly
