create-prediction-market-agent

Scaffold a prediction market agent project in seconds. Picks the right SimpleFunctions wrapper package for your stack — LangChain, OpenAI Agents SDK, or vanilla TypeScript — and generates a working agent.ts you can npm start right away.

npx create-prediction-market-agent my-agent
# or non-interactively
npx create-prediction-market-agent my-agent --framework langchain

---

What it does

Generates a project directory with:

• package.json — pinned to the right wrapper package
• src/agent.ts — a runnable starter agent
• tsconfig.json — strict-mode ES2022
• .env.example — for your OPENAI_API_KEY
• README.md — short usage doc

Frameworks

--framework	What you get
vanilla (default)	Pure TypeScript using prediction-market-context. Smallest dependency footprint.
openai	OpenAI function-calling loop using openai-agents-prediction-markets.
langchain	LangGraph ReAct agent using langchain-prediction-markets.

If you don't pass --framework, the CLI prompts interactively.

Generated agent — vanilla example

import { world, edges, market } from 'prediction-market-context'

async function main() {
  const state = await world()
  console.log('Uncertainty:', state.index.uncertainty)
  console.log('Regime:', state.regimeSummary)

  const { edges: edgeList } = await edges()
  for (const e of edgeList.slice(0, 3)) {
    console.log(`Edge: ${e.title} (${e.executableEdge}c)`)
  }
}

main().catch(console.error)

Generated agent — OpenAI example

import OpenAI from 'openai'
import { predictionMarketFunctions, handleFunctionCall } from 'openai-agents-prediction-markets'

const openai = new OpenAI()
// ... full chat-completions loop with tool dispatch

Generated agent — LangChain example

import { ChatOpenAI } from '@langchain/openai'
import { createReactAgent } from '@langchain/langgraph/prebuilt'
import { predictionMarketTools } from 'langchain-prediction-markets'

const agent = createReactAgent({
  llm: new ChatOpenAI({ model: 'gpt-4o' }),
  tools: predictionMarketTools(),
})

After scaffolding

cd my-agent
npm install
cp .env.example .env  # add OPENAI_API_KEY
npm start

Bugfix vs older releases

Versions before 2.1.0 had two issues with the generated vanilla template:

1. The console.log statement used escaped backticks (\\\``) inside the template literal, which produced malformed console.log(`Edge: ${...}`)` output that crashed at parse time.
2. An orphan top-level index.mjs carried a separate scaffolder implementation referencing the dead /api/public/market/{ticker} endpoint.

Both are fixed: the generated agent.ts is valid TypeScript, and the orphan file is gone.

Sister packages

This scaffolder is the entry point — the wrapper packages do the actual work. If you already know which framework you want, use the wrapper directly:

Stack	Package
Vanilla TS / agnostic	prediction-market-context
OpenAI Agents SDK	openai-agents-prediction-markets
LangChain / LangGraph	langchain-prediction-markets
Vercel AI SDK	vercel-ai-prediction-markets
CrewAI (Python)	crewai-prediction-markets
MCP / Claude / Cursor	simplefunctions-cli, prediction-market-mcp-example

Testing

npm test

14 unit tests covering every template's structure (no calls to dead endpoints, JSON validity for the package.json generators, regression test for the escaped-backtick bug).

License

MIT — built by SimpleFunctions.