Codex + Riggr

Codex is OpenAI’s coding agent - the build path for ChatGPT users. Use it in VS Code (Codex extension), the ChatGPT desktop app (Codex view), or the Codex CLI. Add Riggr once in ~/.codex/config.toml, authorize, then run Riggr Autopilot so the Next.js app lands on your machine.

Came from ChatGPT? Keep web chat for store Q&A; use this page to scaffold the app.

How Riggr works in Codex

| Piece | Role | | --- | --- | | Codex | Plans, writes code, runs commands in your project folder | | Riggr MCP | Read-only Shopify data + project/prompt tools (while you build) | | Riggr API key .env | Powers the finished Next.js app at runtime (not MCP inside the app) |

ChatGPT web Developer mode and Codex are different clients. Web chat uses OpenAI’s cloud connector UI. Codex on your machine uses ~/.codex/config.toml and can see Downloads / your project folder - required for Autopilot.

Quick start (2 minutes)

Recommended if you have the Codex CLI (npm i -g @openai/codex or your preferred install):

codex mcp add riggr --url https://www.hiriggr.com/api/mcp
codex mcp login riggr

Or edit ~/.codex/config.toml (macOS/Linux) / %USERPROFILE%\.codex\config.toml (Windows):

[mcp_servers.riggr]
url = "https://www.hiriggr.com/api/mcp"

Then run codex mcp login riggr if OAuth did not start on add.

Open a project in VS Code + Codex or ChatGPT desktop → Codex, and ask:

Riggr, what are my last 5 orders?

Approve OAuth if prompted. You should get live store data.

If that worked, skip to Riggr Autopilot. If not, use the walkthrough below.

Step-by-step setup

1. 1

   Install Codex where you code

   Pick at least one:

   • VS Code - Extensions → search Codex by OpenAI (openai.chatgpt), install, open the Codex sidebar, sign in with ChatGPT
   • ChatGPT desktop - Install the app, open the Codex view, add or open a local project folder
   • CLI - Install Codex CLI, run codex from a terminal in your project
2. 2

   Confirm Riggr MCP is active

   At hiriggr.com/dashboard your account should show Riggr MCP · Active.

3. 3

   Register Riggr in Codex MCP config

   CLI (writes the same config.toml the IDE and desktop use):

   codex mcp add riggr --url https://www.hiriggr.com/api/mcp

   Or paste into ~/.codex/config.toml:

   [mcp_servers.riggr]
   url = "https://www.hiriggr.com/api/mcp"

   Common mistake: https://www.hiriggr.com/mcp (missing /api). Prefer user-level config so every project sees Riggr; use .codex/config.toml in a trusted project only if you want it scoped.

4. 4

   Complete OAuth

   Run:

   codex mcp login riggr

   Or trigger any Riggr tool from Codex chat and finish Authorize in the browser. Confirm with codex mcp list and under /connections.

5. 5

   Verify in a Codex session

   In VS Code Codex, ChatGPT desktop Codex, or the CLI TUI, ask:

   What MCP servers do you have? List Riggr tools.

First tool call

Riggr, list the last 5 orders from my Shopify store.

1. 1

   Codex calls a Riggr tool

   You should see a tool use for something like list_orders. Approve if prompted.

2. 2

   OAuth if needed

   Browser opens hiriggr.com/oauth/authorize. Authorize once. Token lasts about a year unless you revoke it.

3. 3

   Codex answers with live data

   Order ids, totals, statuses - from your store.

Verify it's working

1. hiriggr.com/connections shows an active Codex / OpenAI-related MCP client
2. codex mcp list shows riggr with URL https://www.hiriggr.com/api/mcp
3. A Codex chat in a project folder answers a live order/product question

Riggr Autopilot

Build with Riggr Autopilot - Codex loads the Spark over the Riggr MCP bridge and finds your API key on disk. You don’t paste a 30KB prompt file.

1. 1

   Open a fresh project folder in Codex

   One app per directory (e.g. ~/Code/riggr-returns). Use VS Code + Codex sidebar, ChatGPT desktop Codex, or codex in that folder.

2. 2

   Copy Talk to your AI

   Open the project in Riggr. Under Talk to your AI, use the emerald Do this first card — copy the kickoff prompt and paste it into Codex. That loads the Spark (or Compose for two Sparks) over MCP. After Add-ons unlock, merge them one at a time from the Sparks list.

3. 3

   Or kick off in plain language

   If you prefer typing it yourself:

   Riggr Autopilot: open my Returns Triage project, load the starter prompt over MCP, find my downloaded Riggr API key .env on this computer, and build the app.

4. 4

   Let Codex work

   It should call Riggr MCP (use_prompt_in_project / riggr://prompts/...), find riggr-*-api-key.env, configure .env.local, and scaffold Next.js. First builds often take 15-30 minutes.

5. 5

   Run the app locally

   Follow Codex’s npm install / npm run dev steps and open the localhost URL.

Details: Setup guide → Riggr Autopilot.

ChatGPT vs Codex

| | ChatGPT (web / Developer mode) | Codex (this page) | | --- | --- | --- | | Job | Explore store + plan the app | Scaffold and run the app on disk | | MCP config | Developer mode / Plugins connector | ~/.codex/config.toml | | Filesystem | No project folder / Downloads access | Yes - Autopilot can find the API key | | Keep using | Ops questions, prompt choice | Everyday build loop |

You can keep both connected. Many merchants explore in ChatGPT, then build in Codex.

Troubleshooting

More: Troubleshooting.

Related pages

• ChatGPT - explore your store in chat, then come here to build
• AI Tools
• Your computer setup - Terminal, Node, npm install, npm run dev
• Cursor - alternate IDE Autopilot path
• Claude Code - Anthropic terminal Autopilot path
• Setup guide - project + API key download
• Sparks
• Troubleshooting