Skip to content
TriedPrompts

More in Build apps & tools.

Connect grocery receipt data to ChatGPT through a custom MCP server

PartialEvidence: Described by posterPosted Last checked

Mike Houston got grocery receipt data into ChatGPT through a custom MCP server, but it only partly went smoothly; setup needed fixes.

How to use this

  1. Tap Copy prompt.
  2. Open ChatGPT.
  3. Paste it into the message box.
  4. Send it, then read the answer before acting on it.
New to this? What's a prompt?

A prompt is the message you send to the AI tool. Copy the one on this page, put your own details where it asks, and read the answer before you act on it.

Prompt

Recreated prompt

The source didn't share its exact prompt, so we wrote one from what it describes. The result shown is the source's, not ours.

What the prompt labels mean

The person didn't share their exact words, so we wrote this from what they described. We haven't tested it, and they didn't use this exact version.

I have my online grocery receipts parsed into a DynamoDB database behind a TypeScript app on AWS Lambda. Help me add an MCP server to the Lambda with tools to list receipts, get a receipt, get price history, get frequently bought items and get spending analysis, so I can connect it to ChatGPT and ask questions about my shopping history.

Try this in:

ChatGPT Used in sourceGemini Grok Claude Used in source

Sends the Recreated prompt shown above. ChatGPT and Grok open with it filled in, and ChatGPT may send it straight away. Gemini and Claude open empty, so we copy the prompt for you to paste. Every button also copies it.

Opens in a new tab. We don't see what you send.

These links are only a shortcut. We're not affiliated with OpenAI, Google, Anthropic or xAI, and your results may not match the source.

Optional stronger version

Our rewrite with extra checks. The poster didn't test it.

Before you start, ask me 1–3 short questions about the details below and wait for my answers. Once I've answered, confirm the plan in one line and run the whole task start to finish.

Goal: Let me ask an AI chatbot questions about my own grocery purchase history.

Questions to ask me first:
- Where is the receipt data now (PDFs, email, spreadsheet or a database)?
- Do you need a one-off analysis or something you can query every week?
- Which chatbot and plan are you using?

Inputs (fill in or confirm):
- Data location: [PDF FOLDER / DATABASE / SPREADSHEET]
- Chatbot and plan: [TOOL AND PLAN]
- Questions I want answered: [QUESTIONS]
- Hosting, if building a connector: [HOSTING]

Rules:
- Suggest the simplest route first, such as uploading files, before building a custom connector.
- If a connector is needed, check the chatbot's current documentation for supported transports, sign-in and plan requirements.
- Don't rely on memory for fast-changing specs; say when advice may be out of date.
- Keep my receipts private and don't send them anywhere I haven't approved.

Checkpoint: Tell me the route you recommend and the plan or account it needs, and wait for my OK.

If you get stuck: If the connection shows as connected but nothing works, list the logs or settings to check one at a time.

Verify before you finish: Ask one test question and confirm the answer matches the receipts.

Output: Setup steps or code, plus a few example questions to ask.

Done when: I can ask questions about my grocery history and get answers that match my receipts.

AI tools

ChatGPTClaude

Outcome

Mike Houston's side project Crosscheckout collects online grocery receipts by email, parses the PDFs and stores line items in a database on AWS Lambda and DynamoDB. Houston wanted to chat with that shopping history, so Houston added an MCP endpoint with five tools, prototyped with Claude Opus in about an hour. Cursor connected easily. ChatGPT did not: it needed OAuth discovery endpoints, said "Connected" but showed nothing, and kept giving outdated advice about streaming transports. Houston ignored that advice, added streaming support, then spent time on empty responses caused by stale API Gateway deploys. The final blocker was account tier: custom MCP servers in ChatGPT need a Plus or Pro plan, though the free settings didn't say so. After upgrading and roughly another hour of testing, the app stayed connected and Houston could ask about apricot stock levels. ChatGPT also pointed out that the data had no pasta or rice because those were bought in person, a sampling bias. Houston admits that just uploading the PDFs would have been far easier if the goal was a one-off analysis.

Evidence: Described by posterThe person describes what happened in words, without screenshots or video.

Known limitations

PartialThe source says it only partly worked, or had a significant problem a person had to deal with.

What went wrong: It worked in the end, but ChatGPT gave outdated transport advice, needed OAuth fixes and a Plus plan, and Houston says uploading PDFs would have been easier.

Source

Posted
by Mike Houston (kothar.net)
Last checked
Method:
How we verify

https://kothar.net/blog/2026/mcp-grocery-receipts

Context: The exact prompt isn't quoted, so the prompt here is recreated. Claude Opus helped write the server; ChatGPT was the chat client. This is a developer project, not a beginner setup.

More in Build apps & tools

Similar cases