Luchamon Bot

A bot that never wastes an energy

Luchamon Bot automates battles on luchamon.com, a fully on-chain auto-battler on the Base chain. Each Luchamon gets seven energies per week; the bot spends them on the best move, never losing one.

It signs its transactions with the game's Fast Mode Wallet — the main wallet stays cold.

Pick by simulation, not heuristics

For each fight, the game offers eight random opponents. Rather than guess, the bot runs each of the eight through the game's own battle engine, over many VRF seeds, to estimate a win probability. It then plays the best available one — the "least bad," since even the loser earns $LUCHA, so no energy is wasted. The objective is tunable per Luchamon: safe aims for the surest win, aggressive for the strongest still-safe opponent to climb the ladder.

These simulations need thousands of on-chain reads. A local Anvil fork serves the same bytecode about 24× faster than the free RPC tier — a calibration run drops from ~40 min to under a minute once the cache is warm.

A run is one drain

npm start is one run: it drains the week's available energy, then exits. The process is the run — it never loops or schedules itself. An always-on core service reads the weekly reset on-chain and launches the drain at the right moment, with progress and a debrief pushed to Telegram.

Local cockpit

A web cockpit (React + Vite, strict TypeScript) drives the bot locally: run an anti-meta analysis, inspect the roster, and apply a build by signing the updateBuild transaction with your main wallet in Rabby or MetaMask. The core service never holds the main key — it prepares the calldata and journals, nothing more. In the Docker stack, nginx serves the bundle and re-injects the API secret server-side: the JS bundle never sees it.

Stack

Built in TypeScript with viem for on-chain interaction, deployable via Docker (core + Telegram + cockpit + Anvil) in one docker compose up. Architectural decisions are recorded as ADRs in a private vault — 24 to date.