ChatGPT Plus users can install Custom GPTs that call external APIs via OpenAI Actions. With a market data API like TradeOdds, you can build a Custom GPT that runs real historical analyses, ranked universe scans, and explains the results — instead of guessing or hallucinating numbers.

This guide walks through building a stock-analysis Custom GPT in roughly ten minutes.

Prerequisites

ChatGPT Plus, Team, or Enterprise (Custom GPTs require a paid ChatGPT plan).
A TradeOdds account with Pro tier or above ($29/month).
Your TradeOdds API key from Account → API Keys.

Step 1: Download the TradeOdds Custom GPT config

Go to tradeodds.io/api-docs and find the OpenAI Custom GPT integration card. Download the Actions JSON file. The config contains the OpenAPI spec for TradeOdds’ analyze, factor_match, query, and list_symbols endpoints — pre-formatted for OpenAI Actions.

Step 2: Create a new Custom GPT in ChatGPT

In ChatGPT, click your profile → My GPTs → Create a GPT. Switch to the Configure tab (skip the conversational builder).

Step 3: Fill in basic info

Name. Something like “TradeOdds Stock Analyst” or “Historical Base Rates.”
Description. “Looks up historical base rates and runs universe-wide scans against 35 years of clean market data.”
Instructions. Use the following as a starting point — adjust for your style:

You are a quantitative stock analyst with access to TradeOdds, a 35-year
historical market database with 16 pre-computed factor columns (RSI zones,
VIX buckets, market regime, streaks, magnitude, earnings proximity, etc.).

When the user asks about a stock's historical behavior, use the `analyze`
tool. When they ask "which stocks have X conditions today," use
`factor_match`. When they ask a question requiring custom SQL (Power User
tier required), use `query`.

Always show the sample size and the date range. Never present base rates
as predictions — they are historical frequencies. If a user asks for a
trade recommendation, decline and instead show the historical record.

The user's TradeOdds tier may limit which tools are available. If a tool
returns a 403, explain that the feature requires an upgrade.

Conversation starters. Add 4 prompts to help users discover use cases. Examples:
- “What’s SPY’s historical 5-day win rate after a 3%+ drop with VIX above 25?”
- “Find me the top 10 stocks in a bull regime with oversold RSI today.”
- “How often does QQQ recover from a 2% gap down within 5 days?”
- “List symbols covered by TradeOdds.”

Step 4: Add the Actions

In the Configure tab, scroll to Actions → Create new action. Paste the OpenAPI spec from the downloaded JSON. The fields auto-populate.

Under Authentication, select API Key. Set:

Auth Type: Bearer
API Key: Your sk-to-... key from TradeOdds

ChatGPT will now send Authorization: Bearer sk-to-your-key with every Action call.

Step 5: Test before publishing

In the right-hand preview pane, ask: “What’s the historical 5-day win rate for SPY when it drops 3% with VIX above 25?”

The Custom GPT should call the analyze action, return real numbers, and present them with proper sample-size context. If it returns a 401 or 403, recheck the API key.

Step 6: Publish

Click Save in the top-right. Choose visibility:

Only me — private, just for you.
Anyone with the link — shareable.
Public — listed in the GPT Store.

For a personal trading assistant, “Only me” is the right choice. For a team tool, share via link. The GPT Store option exposes you to a broader audience but requires more polish.

What this Custom GPT can do

Once active, the GPT lets you ask natural-language questions and gets real answers:

“What does the historical record say about NVDA after a 5%+ rally on heavy volume?”
“Find me stocks above $20 in bear regime with RSI oversold and VIX above 30.”
“Run a SQL query to count every SPY day in the last 5 years where the next-day return was above 2%.”

The GPT shows the underlying conditions and sample size so you can verify any result.

Comparison: Custom GPT vs Claude MCP

	OpenAI Custom GPT	Claude MCP
Where it runs	ChatGPT (web/app)	Claude Desktop (or Cursor)
Auth	API key in OpenAI Actions config	API key in local config file
Sharing	Public/link via GPT Store	Per-machine config
Persistence	Tied to your ChatGPT account	Tied to your local install
Best for	ChatGPT-centric workflows	Desktop / code-editor workflows

If you use both ChatGPT and Claude, you can install both — they use the same TradeOdds API key under the hood.

Troubleshooting

Action returns 401. Your API key was rejected. Verify it’s correct, has the sk-to- prefix, and isn’t revoked.

Action returns 403. The endpoint requires a higher subscription tier. Upgrade or scope the GPT’s instructions to avoid that endpoint.

GPT hallucinates instead of calling the action. Strengthen the instructions: “Always use the analyze action when the user asks about a stock’s historical behavior. Never answer from memory.”

Slow responses. Custom GPT actions add a round-trip; expect ~2 seconds per tool call. Multi-tool calls add up.

What’s next

Once the Custom GPT is working, common upgrades:

Add custom instructions that reference your own trading style or risk parameters.
Combine TradeOdds with another data source (news, fundamentals) via additional Actions.
Share the GPT with your trading group so everyone references the same base rates.

Verification. Setup steps verified on June 4, 2026 against OpenAI’s Custom GPT Actions feature and TradeOdds’ published OpenAPI 3.1 spec. The Custom GPT builder UI evolves frequently; check ChatGPT’s docs if specific UI steps don’t match.

Disclaimer. TradeOdds provides historical analysis for informational purposes only. This is not investment advice. Past performance does not guarantee future results.

How to Build a Custom GPT for Stock Analysis (2026 Guide)