mcpField
A
Another™ MCP Server for Binary Ninja

Another™ MCP Server for Binary Ninja

Local
github.com/MCPPhalanxSecurity51Updated 76w ago

About Another™ MCP Server for Binary Ninja

Another MCP Server for Binary Ninja is an MCP server that exposes binaries opened in Binary Ninja to AI clients for reverse engineering. It runs either as a Binary Ninja UI plugin or headless from the command line, and is aimed at security researchers and reverse engineers.

Key features

  1. 01Rename functions and data variables
  2. 02Retrieve pseudo C, pseudo Rust and IL views
  3. 03Inspect disassembly, imports, exports and sections
  4. 04List strings, functions and data variables
  5. 05Connect via SSE or stdio transport

Use cases

  1. 01Asking an AI client to explain decompiled functions
  2. 02Serving binary analysis on a headless analysis host
  3. 03Querying binary structure from Claude Desktop or Cherry Studio

MCP support

This project is an MCP server that exposes each file opened in Binary Ninja as a resource under binaryninja:// URIs and provides tools such as pseudo_c and disassembly for MCP clients.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

mcp.json · Universal config
{
"mcpServers": {
"binaryninja": {
"command": "uvx",
"args": [
"binaryninja-mcp"
]
}
}
}
Run commanduvx binaryninja-mcp

Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.

Packages
pypibinaryninja-mcp

Tools

No tool list yet

Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.

Details

From the project README

Another™ MCP Server for Binary Ninja

The MCP (Model Context Protocol) Server for Binary Ninja

PyPI Apache licensed Python Version GitHub Discussions

English | 中文

Demo

The tests/binary/beleaf.elf is taken from CSAW'19: Beleaf - Nightmare. You can also find the complete writeup from the link above!

demo

... but why Another?

See: Key Differences from the Existing Plugin

Installation

Server Setup

There are two ways to run the MCP server:

  1. Binary Ninja UI Plugin:

    • Install the plugin via Binary Ninja's plugin manager
    • The MCP server will start automatically when first file is loaded.
      • Auto start is configurable via Settings - MCP Server - Auto Start
      • Listen port is configurable via Settings - MCP Server - Server port number
    • All opened files are exposed to separate resources, see Available Resources section below
  2. Binary Ninja Headless Mode:

    uvx binaryninja-mcp install-api  # only run once
    uvx binaryninja-mcp server <filename> [filename]...
    
    • filename could be any binary files or BNDB, like in UI mode, all opened files are available to the MCP client.
    • Server runs on default port 7000
    • Use --port flag to specify a different port

MCP Client Setup

  1. Claude Desktop (stdio relay client): Configure the client to connect via stdio transport using built-in relay.

    {
      "mcpServers": {
        "binaryninja": {
          "command": "uvx",
          "args": ["binaryninja-mcp", "client"]
        }
      }
    }
    
  2. Cherry Studio:

    • SSE endpoint (recommanded): URL: http://localhost:7000/sse
    • stdio client:
      • Command: uvx
      • Arguments:
        binaryninja-mcp
        client
        

Add --port 12345 to both server and client command line if you prefer to run MCP server on port other than default.

Available Tools for MCP Clients

The MCP server provides the following tools:

  • rename_symbol: Rename a function or a data variable
  • pseudo_c: Get pseudo C code of a specified function
  • pseudo_rust: Get pseudo Rust code of a specified function
  • high_level_il: Get high level IL of a specified function
  • medium_level_il: Get medium level IL of a specified function
  • disassembly: Get disassembly of a function or specified range
  • update_analysis_and_wait: Update analysis for the binary and wait for completion
  • get_triage_summary: Get basic information from BinaryNinja Triage view
  • get_imports: Get dictionary of imported symbols
  • get_exports: Get dictionary of exported symbols
  • get_segments: Get list of memory segments
  • get_sections: Get list of binary sections
  • get_strings: Get list of strings found in the binary
  • get_functions: Get list of functions
  • get_data_variables: Get list of data variables

Available Resources for MCP Clients

MCP Resources can be accessed via URIs in the format: binaryninja://{filename}/{resource_type}

The server provides these resource types for each binary:

  • triage_summary: Basic information from BinaryNinja Triage view
  • imports: Dictionary of imported symbols/functions
  • exports: Dictionary of exported symbols/functions
  • segments: List of memory segments
  • sections: List of binary sections
  • strings: List of strings found in the binary
  • functions: List of functions
  • data_variables: List of data variables

Development

uv is the recommanded package management tool for this project.

Clone directory to Binary Ninja Plugin Directory

git clone https://github.com/MCPPhalanx/binaryninja-mcp.git "${env:APPDATA}\Binary Ninja\plugins\MCPPhalanx_binaryninja_mcp"

Setup Python Environment

Binary Ninja API must be installed into virtualenv manually.

uv venv
uv sync --dev
# install binaryninja API
binaryninja-mcp install-api
# check API is correctly installed
uv run python -c 'import binaryninja as bn; assert bn._init_plugins() is None; assert bn.core_ui_enabled() is not None; print("BN API check PASSED!!")'

Setup MCP Client for Development

For MCP clients with stdio transport like Claude Desktop, change working directory to development folder.

{
  "mcpServers": {
    "binaryninja": {
      "command": "uv",
      "args": [
        "--directory",
        "C:/path/to/binaryninja-mcp",
        "run",
        "binaryninja-mcp",
        "client"
      ]
    }
  }
}

SSE-enabled MCP clients can connect using: http://localhost:7000/sse

Build

uv build

Test

pytest
# To update test snapshots:
pytest --snapshot-update

Version Bump

The PyPI package version is automatically derived from Binary Ninja's plugin.json (using package.json format), maintaining version consistency between the BN plugin and PyPI package.

# bump alpha version
uvx hatch version a

# bump release version
uvx hatch version minor,rc
uvx hatch version release

See: Versioning - Hatch

Release

uv publish

License

Apache 2.0

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.