Skip to main content

​
Overview

ChatGPT doesn’t use MCP — it uses GPT Actions, which are powered by OpenAPI specs. The good news: EventGraph already has a complete OpenAPI spec, so you can create a Custom GPT in minutes with no code.

​
Setup

​
1. Go to ChatGPT → Explore GPTs → Create

  1. Open chat.openai.com
  2. Click Explore GPTs in the left sidebar
  3. Click Create (top right)

​
2. Configure the GPT

In the Configure tab: Name: EventGraph — Prediction Market Intelligence Description: 
I have access to live prediction market data from Polymarket, Kalshi, Limitless, 
and OpinionTrade via the EventGraph API. Ask me about active markets, prices, 
volumes, arbitrage opportunities, and more.
Instructions: 
You are a prediction market analyst with real-time access to EventGraph data.

When the user asks about prediction markets, events, or probabilities:
1. Use the search_markets or search_events action to find relevant markets
2. Use get_market for detailed data on a specific market
3. Use find_arbitrage to identify pricing discrepancies across platforms (Pro)
4. Always cite the platform, current price, and volume when discussing a market
5. Express prices as probabilities (e.g. "67% chance of Yes" not "0.67")

Be concise, data-driven, and helpful.

​
3. Add the EventGraph Action

In the Configure tab, scroll to Actions and click Create new action. Authentication: Select API Key, header name: X-API-Key, value: eg_live_your_key_here OpenAPI schema: Paste this: 
openapi: 3.1.0
info:
  title: EventGraph API
  description: Live prediction market data from Polymarket, Kalshi, Limitless, and OpinionTrade
  version: 1.0.0
servers:
  - url: https://app.eventgraph.ai
paths:
  /api/v1/markets:
    get:
      operationId: search_markets
      summary: Search prediction markets
      parameters:
        - name: query
          in: query
          schema:
            type: string
          description: Keyword search query
        - name: platform
          in: query
          schema:
            type: string
            enum: [polymarket, kalshi, limitless, opiniontrade]
        - name: status
          in: query
          schema:
            type: string
            enum: [active, resolved, all]
            default: active
        - name: page_size
          in: query
          schema:
            type: integer
            default: 20
      responses:
        "200":
          description: List of markets
  /api/v1/markets/{market_id}:
    get:
      operationId: get_market
      summary: Get market details
      parameters:
        - name: market_id
          in: path
          required: true
          schema:
            type: string
      responses:
        "200":
          description: Market details
  /api/v1/events:
    get:
      operationId: search_events
      summary: Search events
      parameters:
        - name: query
          in: query
          schema:
            type: string
        - name: category
          in: query
          schema:
            type: string
        - name: page_size
          in: query
          schema:
            type: integer
            default: 20
      responses:
        "200":
          description: List of events
  /api/v1/arbitrage:
    get:
      operationId: find_arbitrage
      summary: Find cross-platform arbitrage opportunities (Pro)
      parameters:
        - name: min_spread
          in: query
          schema:
            type: number
        - name: page_size
          in: query
          schema:
            type: integer
            default: 20
      responses:
        "200":
          description: Arbitrage opportunities
  /api/v1/status:
    get:
      operationId: get_status
      summary: API health and data freshness
      responses:
        "200":
          description: Status

​
4. Save and test

Click Save, then test in the preview panel:
“What are the most traded prediction markets this week?”

​
What it looks like

Once live, your Custom GPT can answer questions like:
QuestionWhat it does
”What’s the current probability of a Fed rate cut in June?”Calls search_markets with query, returns live Polymarket/Kalshi prices
”Are there any arbitrage opportunities right now?”Calls find_arbitrage, lists spreads (Pro)
“Compare all platforms on the 2026 election market”Calls search_events then get_market for each result

​
Share your GPT

Once tested, you can publish your Custom GPT publicly so others can use it — no API key management needed on their end (your key is embedded in the Action config, not visible to users).
If you publish publicly, your EventGraph API key is used for all queries. Monitor your usage at GET /api/v1/usage to avoid exceeding limits. Consider upgrading to Pro for higher rate limits before going public.

​
OpenAI Assistant API (for developers)

If you’re building programmatically, you can define EventGraph as a function tool in the Assistants API: 
tools = [
    {
        "type": "function",
        "function": {
            "name": "search_markets",
            "description": "Search prediction markets across Polymarket, Kalshi, Limitless, and OpinionTrade",
            "parameters": {
                "type": "object",
                "properties": {
                    "query": {"type": "string", "description": "Keyword search"},
                    "platform": {"type": "string", "enum": ["polymarket", "kalshi", "limitless", "opiniontrade"]},
                    "page_size": {"type": "integer", "default": 20}
                }
            }
        }
    }
]
Then handle the tool call by proxying to https://app.eventgraph.ai/api/v1/markets with your X-API-Key header.