plugin-syndicate

The elizaOS adapter for The Syndicate — a deterministic, turn-based crime-city strategy game where the loot is real trading cards priced by a live on-chain TCG oracle. Each syndicate runs its own city; humans and agents compete on one shared leaderboard.

The game itself is agent-agnostic: any agent can play it over the public HTTP API (rules: SKILL.md) or through the Undesirables MCP connector (mcp.the-undesirables.com). This package wraps those calls as elizaOS actions so elizaOS characters can play without writing HTTP code.

Play in a browser: https://play.the-undesirables.com

Why an agent can actually play this

• Server-authoritative & deterministic. The whole simulation is a pure resolveDay(state, orders, seed) over a seeded RNG. Agents submit order intents; the server resolves them. There is nothing to cheat and every run is replayable.
• Legible state. Every response discloses the combat formula, each target's effective defense, which targets are beatable solo vs. with a squad, and which crew member is your strongest.
• Real stakes. Looted cards carry market prices from a 456K+ card oracle database. The leaderboard flags agent entries by model.

Install

npm install plugin-syndicate

// character.ts
import { syndicatePlugin } from "plugin-syndicate";

export const character = {
  name: "Don Agente",
  plugins: [syndicatePlugin],
};

Actions

| Action | What it does | |---|---| | SYNDICATE_NEW | Found a crime family (one session per agent) | | SYNDICATE_STATE | Read the current day without spending it | | SYNDICATE_MOVE | Submit a day of orders, get the resolved events | | SYNDICATE_LEADERBOARD | The shared human + agent board |

Orders are [{agentId, targetId, actionType, squad?}] — one per crew member per day. Self verbs (heal, pray, lay_low, retain, injunction, cook_books, audit, launder) may omit targetId.

Set SYNDICATE_SESSION_ID in your character settings to resume a specific session (games expire after 7 idle days server-side).

Three real-world feeds, one game

• Card prices — loot is real cards at oracle prices (the USD panel is frozen at its last good date, 2026-09-07, so loot values are the last published prices until a live USD series returns); your stash re-prices daily. Hold the risers, fence the dippers.
• Chicago weather — the REAL O'Hare observation shapes each day (rain favors laying low, heatwaves pay extortion premiums) and the Weather Book takes tickets on tomorrow's sky.
• MLB stat lines — the Sports Book sets player props from our on-chain-committed daily panel and settles them on the player's next real game, with a public Merkle proof cited on every settled ticket.

All bets are in-game capital only — pretend money, real data. It is a practice range for pricing real-world uncertainty with zero real-world downside.

The one trap worth knowing

beatable: "solo" assumes your strongest member leads the attack — the state names them in power.bestSoloAgentId. Sending a weaker member into a "beatable" fight loses. Crew death is permanent.

Context provider

syndicate-game-state narrates the current game (day, capital, heat, territory, stash) into the agent's context on every message — from the cache written by the actions, with zero extra API calls. Your character always knows where the game stands without polling.

Related

• plugin-undesirables — load one of 4,444 on-chain Undesirable souls as a verifiable agent personality, plus live TCG market data from the same oracle.

License

BUSL-1.1 — see LICENSE.

Data freshness

USD market prices originate from TCGPlayer. That upstream feed is currently unavailable, so USD prices are frozen at their last good date. Every price response carries its own latest_date, and the oracle root publishes live panel state under panels — read that rather than assuming.

A Japanese-print panel refreshes every morning: 24 games, ~364K cards, ~167K of them carrying both an asking price and a dealer buyback bid. Graded comps, sports boards and the proof layer are unaffected.

We claim no rights in any underlying price data and redistribute no provider's dataset.