AgenticSTS
Research testbed for long-horizon language-model agents on Slay the Spire 2, built around bounded, typed memory instead of a growing transcript. Reports 3/10 wins at Ascension 0 without learned skills and 6/10 with them.
Every idea for building Slay the Spire tooling, in one place.
Mods, game bridges, simulators, bots, seed tools, run data and research for Slay the Spire and Slay the Spire 2, with notes on what each approach takes.
Most Spire tooling needs the same three pieces: a way to read the game, something that decides or analyzes, and a way to test faster than real time. This index collects what people have built for each.
Start Here
What it takes
The parts every bot needs, from reading game state to choosing a path, and where each part gets hard.
Read →Approaches
Rules, search, reinforcement learning, learning from run data and language models, with the projects that tried each.
Read →Simulators
Headless reimplementations that let a bot try thousands of lines before the real game draws a card.
Read →Run data
Public run histories for training, evaluating and checking an idea against how people actually play.
Read →New to modding?
Most tooling starts as a mod. Slay the Spire mods are Java on ModTheSpire and BaseMod; Slay the Spire 2 mods are C# on the game's official mod support. The modding page has a step-by-step path and the templates, tutorials and decompilers for each game.
Start moddingFeatured Projects
Bottled AI
Full-run bot on Communication Mod using hand-built priority lists for choices and a simulated search over hand orderings in combat. The README reports a 52% win rate with its best Watcher strategy.
Communication Mod
A ModTheSpire mod that launches an external process and talks to it over stdin/stdout, sending game state as JSON and accepting text commands. Most Slay the Spire bots are built on it.
Spire Codex
Database and public API for Slay the Spire 2 (cards, relics, monsters, events and more in 15 languages), built from the game files, plus community run statistics, leaderboards and patch-to-patch diffs.
spirecomm
Python package that wraps Communication Mod's protocol in game-state objects and actions, with a simple example agent.
sts_lightspeed
Standalone C++17 reimplementation of Slay the Spire built for tree search, designed to be RNG-accurate. The README reports 1M random playouts in 5 seconds on 16 threads.
sts-ironclad-agent
A20 Ironclad bot: one frozen network makes every out-of-combat choice and simulator search plays combat. The README reports a 50.1% Heart win rate over 501 runs in the original game (95% interval 45.7–54.5%).
sts-rl-agent
A0 Ironclad agent: a learned policy handles every out-of-combat choice and sts_lightspeed MCTS plays combat. On 50 held-out seeds it reaches floor 42.5 and wins 14%, against 31.2 and 6% for heuristics.
sts2-cli
Runs the real Slay the Spire 2 engine headless in a terminal, so damage, card effects, enemy AI, relics and RNG match the game exactly. It needs a Steam copy of the game, and the README says it was tested on v0.111.0.
STS2FableBot
Full-run Slay the Spire 2 bot with a deterministic, search-based decision engine, playing through a fork of STS2MCP. The README reports a 20.7% win rate at Ascension 0 over 343 runs and 19.4% at Ascension 1 over 175.
STS2MCP
Slay the Spire 2 mod that exposes game state and actions over a localhost REST API, with an optional MCP server so language-model agents can play. Covers single player and co-op (beta).
By Category
7 Game interface Mods and bridges that expose game state and accept actions. 7 Modding foundation Loaders, libraries and docs that every mod, bot bridge included, builds on. 1 Client library Code that speaks a game interface so a bot doesn't have to. 17 Bot / agent Programs that play the game end to end or a part of it. 5 Simulator Reimplementations of the game for fast, headless rollouts. 6 Dataset Run histories and logs to learn from or evaluate against. 12 Community tool Seed finders, trackers, overlays and advisors. 12 Research Papers, theses and write-ups.