A Warcraft 3 Reforged bot written in Python — Orc race, Blademaster focus.
Reads live game state directly from process memory, executes macro and micro faster than any human hand. No machine learning. No cheats. Pure engineering.
Forged in code. Sharpened in battle.
- Memory reader — reads gold, lumber, food, unit positions and building states directly from WC3 process memory via macOS Mach APIs
- Sidecar system — injects a Lua script into the WC3 map that exposes structured game state
- Action queue — priority-based input queue (MICRO > ARMY > MACRO > ECONOMY) with preemption between steps
- Macro loop — automated build order execution (peons, Great Hall, Barracks, tech)
- Micro loop — Blademaster ability usage and unit control
- Input executor — single-threaded input dispatch via CGEvent (keyboard) and CGEventPost (mouse)
- Shorts pipeline — automated YouTube Shorts production from bot sessions
SidecarScanner (memory reads)
│
▼
MacroLoop ──► ActionQueue ──► InputExecutor ──► WC3Input
MicroLoop ──► (single thread)
- macOS (Apple Silicon or Intel)
- Python 3.11+
- Warcraft 3 Reforged (windowed/borderless mode)
- StormLib (
brew install stormlib) for map injection
pip install -r tools/shorts_pipeline/requirements.txtsrc/wc3bot/
action/ input primitives, build order
core/ action queue, executor, macro/micro loops, coords
observe/ memory reader, sidecar scanner, state parser
race/orc/ orc-specific config, setup, build order
scripts/ diagnostic + test scripts
tools/
shorts_pipeline/ automated YouTube Shorts production
maps/sidecar/ Lua sidecar map injection
Currently in Phase 1 — beating WC3 Insane AI consistently with the Orc race.
- Live game state via memory reads
- Sidecar Lua injection
- ISO coordinate system + screen projection
- Threading architecture (ActionQueue + InputExecutor)
- T16 macro start sequence — 0.56s
- Blademaster micro loop
- Win rate ≥ 7/10 vs Insane AI
| 📺 YouTube | @GrimForgeBot — every session recorded, every failure logged |
| 🔥 Ko-fi | ko-fi.com/grimforge — keeps the forge burning |
MIT