STS2MCP

An Experimental Research Project to Fully-Automate your Slay the Spire 2 Runs

A mod for Slay the Spire 2 that lets AI agents play the game. Exposes game state and actions via a localhost REST API, with an optional MCP server for Claude Desktop / Claude Code integration.

Singleplayer and multiplayer (co-op) supported, plus full menu and lobby control: profile switching, character select (SP and MP host/client) with optional seed, multiplayer host / Steam-friend join / FastMP localhost join, multiplayer load lobby for resuming saved co-op runs, game-over dismissal, FTUE/tutorial popup handling, and Timeline visibility. Tested against STS2 v0.103.2.

For Players

1. Install the Mod

Grab the latest release and follow the instructions:

1. Copy STS2_MCP.dll and STS2_MCP.json to <game_install>/mods/
2. Launch the game and enable mods in settings (a consent dialog appears on first launch)
3. The mod starts an HTTP server on localhost:15526 automatically

macOS install

On macOS, the mods directory lives inside the app bundle. The default Steam install path is:

~/Library/Application Support/Steam/steamapps/common/Slay the Spire 2/
    SlayTheSpire2.app/Contents/MacOS/mods/

To install, right-click SlayTheSpire2.app → Show Package Contents, navigate to Contents/MacOS/, and create a mods folder. Or from the terminal:

GAME_DIR="$HOME/Library/Application Support/Steam/steamapps/common/Slay the Spire 2"
MODS_DIR="$GAME_DIR/SlayTheSpire2.app/Contents/MacOS/mods"
mkdir -p "$MODS_DIR"
cp STS2_MCP.dll "$MODS_DIR/"
cp STS2_MCP.json "$MODS_DIR/"

Launch the game and open Settings → Mods. The mod should appear in the list. A consent dialog appears on first launch — accept it to enable mod loading. Once enabled, verify the HTTP server is running:

curl -s http://localhost:15526/

A successful response looks like:

{"message": "Hello from STS2 MCP v0.3.4", "status": "ok"}

If you get “Connection refused”, the mod is not loaded — check that mods are enabled in the game’s settings.

2. Give Your AI Instructions to Interact with the Game

Clone or download the repository, then:

MCP server setup

Install uv if you don’t have it (macOS: brew install uv). Then run the server once to install dependencies:

uv run --directory /path/to/STS2_MCP/mcp python server.py --help

uv reads mcp/pyproject.toml, creates an isolated virtual environment, and installs the pinned dependencies from mcp/uv.lock. Subsequent runs reuse the environment instantly.

Add the server to your AI client’s MCP config:

{
  "mcpServers": {
    "sts2": {
      "command": "uv",
      "args": ["run", "--directory", "/path/to/STS2_MCP/mcp", "python", "server.py"]
    }
  }
}

Claude Code: add to your project’s .mcp.json. Claude Desktop: add to claude_desktop_config.json with the same config as above. Other agents should have similar config options for custom MCP servers.

Restart your Claude session after adding the config. To verify the MCP server is working, ask Claude to call get_game_state — with the game running, it should return the current game state.

The MCP server accepts --host and --port options if you need non-default settings.

Flag --no-trust-env can be used to disable requests from picking up proxy settings from the environment, which can cause connection issues if you are running the server in a container.

Profile and Compendium Data

The HTTP API exposes profile-level progress separately from live run state:

• GET /api/v1/profile returns the active profile’s persistent progress summary, including discoveries, achievements, epochs, character totals, and global run totals.
• GET /api/v1/compendium groups that progress into the same high-level sections as the in-game Compendium: Card Library, Relic Collection, Potion Lab, Bestiary, Character Stats, and Run History.
• GET /api/v1/wiki?query=... searches discovered card and relic wiki entries for the active profile with fuzzy matching. Results are limited to 10 by default and can be overridden with limit; card results include base and upgraded variants when available.
• GET /api/v1/profiles lists the three profile slots and the active profile.
• POST /api/v1/profiles switches or deletes profile slots through the game UI.

The MCP server exposes the same profile data through get_profile(), get_compendium(), search_wiki(query, item_type, limit), list_profiles(), switch_profile(profile_id), and delete_profile(profile_id).

get_compendium() is intended for agents that need durable context outside the current room or current run. It works from the main menu, includes a current_run block while a run is active, and summarizes saved saves/history/*.run files for the active profile. Run history is capped to the 20 most recent files in the response so long-lived profiles do not create unbounded tool output.

search_wiki() is the selective lookup path for durable card and relic text. It never returns the full game catalog: the mod first filters to the active profile’s discovered card and relic IDs, then returns only the best fuzzy matches. Use item_type="card" or item_type="relic" when the query is known, and raise limit only when the default 10 results are not enough.

For Developers

Build & Install

Requires .NET 9 SDK and the base game.

PowerShell (recommended):

# Pass game path directly:
.\build.ps1 -GameDir "D:\SteamLibrary\steamapps\common\Slay the Spire 2"

# Or set it once and forget:
$env:STS2_GAME_DIR = "D:\SteamLibrary\steamapps\common\Slay the Spire 2"
.\build.ps1

The script builds STS2_MCP.dll into out/STS2_MCP/. Copy it along with the manifest JSON to <game_install>/mods/ to install:

out/STS2_MCP/STS2_MCP.dll           ->  <game_install>/mods/STS2_MCP.dll
mod_manifest.json                   ->  <game_install>/mods/STS2_MCP.json

Build instructions for macOS

Install dotnet to compile the mod:

brew install dotnet@9
export DOTNET_ROOT="/opt/homebrew/opt/dotnet@9/libexec"
export PATH="$DOTNET_ROOT:$PATH"

Homebrew installs dotnet@9 as keg-only, so the exports above are required for the current session. Add them to ~/.zshrc to persist across sessions.

Build with dotnet directly (the PowerShell script is Windows-only):

dotnet build STS2_MCP.csproj -c Release -o out/STS2_MCP \
  -p:STS2GameDir="$HOME/Library/Application Support/Steam/steamapps/common/Slay the Spire 2"

On macOS the game ships as an app bundle. The .csproj detects macOS and resolves the data directory to SlayTheSpire2.app/Contents/Resources/data_sts2_macos_arm64 automatically.

The mods directory on macOS lives inside the app bundle at SlayTheSpire2.app/Contents/MacOS/mods/. Finder hides bundle contents by default — to browse it in the GUI, right-click SlayTheSpire2.app → Show Package Contents. Or copy from the terminal:

GAME_DIR="$HOME/Library/Application Support/Steam/steamapps/common/Slay the Spire 2"
MODS_DIR="$GAME_DIR/SlayTheSpire2.app/Contents/MacOS/mods"
mkdir -p "$MODS_DIR"
cp out/STS2_MCP/STS2_MCP.dll "$MODS_DIR/"
cp mod_manifest.json "$MODS_DIR/STS2_MCP.json"

License

MIT

FAQ

Why let the AI play the game for me?

I start building this mod with the hope that I can co-op with an AI player. Singleplayer is originally just built for validation.

You did not answer the question!

First of all, I play lots of games, including service games that has daily/weekly tasks. I really hoped that modern AI could save me from the grind, which, if you have tried one or more of the GUI agents, never really materialized. Let’s face it: modern AI is still pretty bad at gaming because no one cares.

About my intention, as a researcher that loves playing games, the purpose of STS2MCP is to test AI models and agents in a rarely explored (we call it out-of-distribution) domain. Ultimately, this might turn into a benchmark for evaluating the reasoning and decision-making capabilities of different language models.

STS2 is just an example to show how good (or bad) current AI agents are at playing such games. I have no intention to replace human players with AI, and I would definitely rather play STS2 myself as a big fan of the game.

Is this a cheat mod?

It can be, but it doesn’t have to be. The mod itself does not alter the gameplay. It is just an interface that allows external programs to interact with the game. What you do with that interface is up to you.

How many tokens do a run consume?

I evaluated on the Ironclad. Claude Sonnet 4.6 uses slightly more than 8M tokens (counting both input, output and tool responses) for a full run. GPT-5.4 averages 7.34M tokens. Depending on your prompt and model choice, it can be more or less.

Do you have a roadmap for future features?

The project is still too early to have a clear roadmap. My current focus is to make sure the core features are stable and well-documented. However, I am open to suggestions and contributions from the community.

• Solidifying multiplayer features and fixing bugs is a priority
• Add support for in-game communication in multiplayer runs when collaborating with an AI agent
• Self-reflection and learning from past runs to improve future performance
• Benchmarking different models and agents is also on my mind