This tool serves as a headless Binary Ninja server, speaking the Model Context Protocol (MCP) to grant AI agents full access to advanced reverse-engineering features. It allows agents to perform disassembly, manipulate intermediate languages (IL), apply patches, define types, trace cross-references, and more, all without the need for a graphical user interface. Designed for co-location within agent runtimes in sandboxed environments, it enables agents to inspect, refine, and extend analyses incrementally, automating complex reverse engineering tasks while maintaining interactive steering capabilities.
주요 기능
01Includes a fake backend mode for CI and development without a Binary Ninja license.
02Offers stdio and TCP transports for flexible integration.
03145 GitHub stars
04Exposes 181 tools across 36 feature groups for deep reverse-engineering.
05Provides scripting access via `binja.eval` and `binja.call` for custom workflows.
06Supports read-only by default with safe mutation workflows (undo/redo, transactions).
사용 사례
01Developing agent-driven workflows in sandboxed VM/container environments for binary analysis.
02Automating complex reverse engineering tasks by integrating with AI agents.
03Incrementally improving analysis databases by having agents update types, symbols, and metadata.
04Applying patches and iterating safely with undo/redo within an agent-controlled environment.
