Explore more topics
Trading workflow · reviewed October 11, 2026

How to connect ChatGPT to NexusTrade

Connect ChatGPT to NexusTrade, inspect an actual dated Apple filing response and verify the intended account before accessing owned portfolios.

As of 2026-10-11

Jump to a section

TL;DR

Connect ChatGPT web to NexusTrade with its custom MCP plugin and OAuth, then check the connection by reading a dated public SEC filing. Local Codex tasks use separate MCP settings and sign-in. Research, backtesting and portfolio changes are later tasks with their own tool access and review; a successful filing read establishes only the read workflow.

Connect ChatGPT in the intended environment

Documented setup flow for this client
Documented setup flow for this client
Web and local hosts
This is the hosted web path. A local Codex task has its own MCP settings and sign-in; ordinary desktop chats do not inherit that host’s configuration.

On ChatGPT web, open Plugins → Add → Add custom MCP server. Name it NexusTrade, enter https://nexustrade.io/api/mcp and choose OAuth. Review the notice, select I understand and want to continue, then Create as a plugin. Finish NexusTrade sign-in. Open the plugin’s details and choose Try in chat.

Try a public filing read in ChatGPT

Inspect the request

Run the public filing request after connecting. A fresh response may return a later filing; compare its reporting period, filing date and source URL.

Recorded NexusTrade filing reference

NexusTrade MCP sec_edgar returned this listing in Codex on 2026-10-12. Use it as a reference when checking your own connection.

Show the table · 5 rows, 2 columns
Returned fieldRecorded result
CompanyApple Inc.
Form10-Q
Filing date2026-07-31
Period ended2026-06-27
Accession0000320193-26-000020

Choose a separate next task

These guides are independent choices after the connection check. Review any proposed paid research, historical test or portfolio change before authorizing it.

Choose the surface you are using

Connect NexusTrade in the product where you want to work. ChatGPT web uses a cloud plugin; Codex desktop and CLI use a local MCP configuration. Each has its own sign-in.

NexusTrade’s public AAPL filing read was executed on October 11, 2026. It records the read workflow; the ChatGPT client connection, historical tests, paid research and trading actions require their own results.

Desktop: choose ChatGPT web or a local Codex task

ChatGPT web and local Codex tasks use separate connections. For the custom MCP workflow described here, use ChatGPT web. In the Codex desktop app, open Settings, MCP servers and Add server; choose Streamable HTTP, enter the NexusTrade endpoint, Save, Restart and Authenticate. Check /mcp in that local task. Use the separate Codex guide for its CLI setup. Native ChatGPT desktop plugin availability must be checked in that app; its ordinary chats do not inherit local Codex MCP settings.

NexusTrade MCP URL: https://nexustrade.io/api/mcp
Transport: Streamable HTTP
Authentication: OAuth

Check the connection with a public read

Paste the prompt below. A working connection returns an actual SEC filing result with the company, filing form, reporting period, filing date and SEC URL. Check those fields before continuing. A stock summary without a tool result does not verify the connection.

Prompt
Call NexusTrade sec_edgar with action list_filings, ticker AAPL and limit 1. Report the company, filing form, filing date, period and source URL from the tool response. Do not access portfolios, ingest filings, submit paid tasks or make changes.

Turn the brief into rules you can inspect

Ask for the universe, signal, sizing, exit rule, test period, initial capital and benchmark before submitting a historical test. Keep the returned backtest ID, poll the same run and inspect warnings, orders and drawdown. Use a separate paper deployment when you want forward simulation.

Account tools can create or change resources. Approve the specific next step after reviewing it. A conversation about a promising result is not an instruction to activate live trading.

Prompt
Draft a strategy from this research brief. Show the complete rules and test assumptions first. After I approve the configuration, run one historical test and compare it with the stated benchmark. Stop after the result review; do not deploy.

Follow the task through to a returned result

Start with the source brief, then freeze a rule for a historical test. The paper guide shows how to identify the running simulated book, read its history and pause it. Check that your current connection grants the required tools before saving or activating anything.

Review a separate brokerage workflow

Public and Tradier deployment use the documented NexusTrade broker connections. The Robinhood guide uses Claude Code with Robinhood’s external MCP server. A paper book does not transfer into a brokerage account by itself; verify the live account, permissions and execution schedule in the selected guide.

Optional: check the account before owned-record changes

Optional: verify an owned portfolio before changes

Skip this step if you only want to inspect the public filing in the next step. Before reading or changing an owned portfolio, ask for a read-only list, inspect one returned ID with get_portfolio and compare its name, strategies and deployment type. An authenticated empty list is valid for a new account.

What you should see: For owned records, you have the intended account response or an authenticated empty response. Resolve an account mismatch before saving anything. Public filing inspection does not require an owned-portfolio read.

Identity boundary
A server list or public filing read does not verify access to your owned portfolios.
Prompt
Use only NexusTrade read tools. Call fetch_portfolios, then get_portfolio with one returned portfolio ID. Report the ID, name, strategies and deployment type. If the authenticated list is empty, report that result. Do not create, edit, backtest, deploy or place orders.

Continue exploring

Discussion

Sign in or create a free account to join the discussion.

No comments yet.

Aurora · AI research assistant

Try this in Aurora

Edit the request, then open it in Aurora. You choose when to send it.