Cheaper than aidapro.net. Smarter than ChatGPT-per-function.

Reverse-engineer binaries autonomously.

Decompile drives IDA Pro from a server-side engine. It walks the call graph in parallel, names variables, finds bugs, and explains decompiled code — without the paste-each-function dance.

Start free See the demo

$0.50 of LLM credit on the house. No card.

decompile → sess_1747680000 · kimi-k2.6

live

0x14001A30WinMain

$0.018

0x14002C04ParseConfig

$0.011

0x14002F10decrypt_blob

$0.027

0x14003120rc4_xor

$0.004

0x14004A20install_hook

$0.006

0x140052F0sub_140052F0

0x14005844sub_14005844

refine pass · install_hook

Sets a length-disassembled inline hook at
ntdll!NtCreateFile via WriteProcessMemory.
Trampoline at 0x14004A20 stores the original
prologue; resolver passes the FileObject to
log_open(). Likely user-mode rootkit stub.

vuln candidate

unchecked path length → stack overflow in log_open.

// engine

What ChatGPT-per-function can't do.

We didn't build a chat wrapper. Decompile is a server-driven engine that takes an objective and a binary, and writes you a report.

Autonomous engine

Pick a root function, hand Decompile an objective in plain English, walk away. It crawls callees, refines each summary, and stitches a final report.

Parallel callee analysis

Up to 10 worker threads attacking the call graph at once. A 100-function binary finishes in minutes, not the half-hour a serial tool takes.

Live call tree

Watch the analysis fan out in IDA, per-function status, live cost meter. You see exactly what you’re paying for, while you’re paying for it.

Vulnerability finding

Per-function refinement pass flags stack overflows, unchecked indices, double-free shapes, and other classics — and explains why.

Crackme solver

Recursive objective tracking lets the engine actually follow the protection. It’s solved CTF crackmes other AI-RE tools just narrate around.

Open-source-able plugin

The client is a thin Python script — all the heavy logic runs server-side. Ports to Binary Ninja and Ghidra are next.

// see it run

Four minutes. One real binary.

Artie's original demo. Watch the engine fan out across the call graph, find the bug, and write it up.

// pricing

Subscription + credits, no surprises.

Each plan grants you LLM credit every month. Burn through it, top up in $10/$25/$100 packs, or set auto-recharge. Hard wall by default — no runaway bills.

Full pricing →

Free

$0forever

Feel the engine. One welcome credit on us.

Grant $0.50 one-time welcome credit