> ## Documentation Index
> Fetch the complete documentation index at: https://docs.kong.fyi/llms.txt
> Use this file to discover all available pages before exploring further.
# Kong
> The world's first AI reverse engineer — LLM-orchestrated binary analysis that recovers function names, types, and structures from stripped binaries.
# The world's first AI reverse engineer
LLM-orchestrated binary analysis that recovers function names, types, and structures
from stripped binaries — so you don't have to.
Go from install to fully-analyzed binary in under five minutes.
See Kong tear apart the XZ Utils backdoor and recover meaningful symbols from a real-world implant.
***
## What Kong does for you
**Stripped binary** — a compiled executable with all debug symbols (function names, type info, variable names) removed. This is the default for release builds and malware alike. Reversing one manually is slow, tedious work.
Kong plugs an LLM into Ghidra's analysis engine and runs a fully autonomous pipeline: decompile, analyze, rename, retype, and export. One command. No babysitting.
A single command runs the complete analysis from triage to export. No manual intervention required — just point it at a binary and walk away.
Runs Ghidra's analysis engine in-process via PyGhidra. No server, no RPC, no subprocess overhead. Just direct, fast access to the decompiler.
Identifies and removes obfuscation techniques using an LLM-driven symbolic tool loop. Control-flow flattening, opaque predicates — Kong handles it.
A post-analysis pass that unifies naming conventions and synthesizes struct definitions across the entire binary. Consistency, automatically.
Works with Anthropic (Claude) and OpenAI (GPT-4o) out of the box, plus custom endpoints. Bring your own model if you want.
Functions are analyzed bottom-up so callers benefit from already-resolved callee context. Smarter naming, better types, fewer hallucinations.
***
## See it in action
