balatro-agentSAFE
Let AI play Balatro
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
English | 简体中文
[](https://www.npmjs.com/package/balatro-mcp) [](https://nodejs.org) [](https://bun.sh) [](https://www.typescriptlang.org) [](https://modelcontextprotocol.io) [](https://github.com/Steamodded/smods)
Let an AI agent play Balatro. Balatro Agent connects any MCP-compatible AI client to a running game. The agent can read live state, pick blinds, play and discard hands, buy jokers, and perform any action available to a human player. It needs no screen capture or vision model; everything works through text tool calls.
[!IMPORTANT] Balatro v1.1 is coming. I will try to provide support asap when its released. Any release after that will not guarantee compatibility with the v1.0 game.
Requirements
- Balatro on Steam
- Lovely Injector
- Steamodded (SMODS)
- Node.js 20 or later for npx (it launches a native MCP binary; Bun, cloning, and building are not required). Native archives from GitHub Releases also run without Node.
Install
1. Install Lovely and SMODS
Follow the SMODS installation guide for your platform. This step is the same for every Balatro mod.
#
cb8e44048bb2OBSERVED · 2026-10-10Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control.
claude mcp add balatro-mcp -- npx -y [email protected]
{
"mcpServers": {
"balatro-mcp": {
"command": "npx",
"args": [
"-y",
"[email protected]"
]
}
}
}Exposed tools (21)
19 read · 2 write · 0 destructive.
| Tool | Risk | Description |
|---|---|---|
ante | write | Ante overview readable throughout a run: the Small, Big, and Boss blinds with chip targets, skip rewards, on-deck blind, boss reroll cost, and queued tags. |
balatro_buy_consumable | read | |
balatro_new_game | read | |
balatro_reorder_hand | read | |
balatro_reorder_jokers | read | |
balatro_select_booster_card | read | |
balatro_select_hand_cards | read | |
balatro_sort_hand | read | |
balatro_use_consumable | read | |
balatro_wiki_search | read | |
booster | read | Currently open booster pack: kind, picks remaining, and options. Errors UNAVAILABLE when no pack is open. |
connect | read | |
consumables | read | Held Tarot, Planet, and Spectral cards with usability status. |
deck | read | Balatro-style Remaining and Full Deck views with base/effective rank, suit, and card-type tallies; face-down remaining cards count as unknown. |
disconnect | read | |
hand | read | Current hand cards with card IDs, modifiers, and selection state, listed left to right in play/scoring order; face-down cards are hidden. |
jokers | read | Owned jokers with editions, costs, and live effect descriptions; face-down jokers are hidden. |
new_postgame | read | |
run | write | Run-level facts: ante, money, poker-hand levels with play counts, vouchers, queued tags, discard pile, challenge, disabled entities, and endless mode. |
shop | read | Shop contents while the shop is open: cards, vouchers, boosters, and reroll cost. Errors UNAVAILABLE outside the SHOP phase. |
turn | read | Live turn snapshot: phase, ante, money, round progress, legal actions, hand with selected cards, jokers, and consumables. |
Trust audit
SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | PASS |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | PASS |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (10 observation(s))
- Network
- declared (2 observation(s))
- Shell
- none-observed
- Dependencies
- not all pinned
- Secrets in source
- none-found
Findings (7)
.oxfmtrc.json
import editionsMarkdown from "../../data/card_modifiers/editions.md" with { type: "text" }import enhancementsMarkdown from "../../data/card_modifiers/enhancements.md" with { type: "text" }import indexMarkdown from "../../data/card_modifiers/index.md" with { type: "text" }import sealsMarkdown from "../../data/card_modifiers/seals.md" with { type: "text" }import stickersMarkdown from "../../data/card_modifiers/stickers.md" with { type: "text" }@modelcontextprotocol/server, rehype-parse, rehype-remark, remark-gfm, remark-stringify, unified, unist-util-visit, zod
Gates applied: no_behavioural_pass.
cb8e44048bb2full audit observations/trust-audit/mcp-server/arcadi4__balatro-agent.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-10 | cb8e44048bb2 | SAFE | B | 89 | first audit |
Questions
What is the balatro-agent MCP server?
Let AI play Balatro
What tools does balatro-agent expose?
21 in total: 19 read-only, 2 that write, and 0 that can delete or overwrite. Every one is listed on this page with its risk.
Is balatro-agent safe to connect to an agent?
The audit found nothing in the source that contradicts what it says it does, and graded it B (89/100). Grade A is held back for packages that have also passed a sandboxed behavioural run, which is why a clean server reads B.
What credentials does balatro-agent need?
No credential environment variables were found in its source, so it appears to need none.
How does balatro-agent run?
It speaks stdio, so it runs as a local process your client starts. It is published on npm as balatro-mcp at 0.2.5.
How current is this page?
The grade is for one exact copy of the source (cb8e44048bb2), read on 2026-10-10. The repository is watched and re-audited when it changes.