Independent project, in development

Lunch Money, in plain language.

lunchmoney.sh is an independent, unofficial project building a Model Context Protocol (MCP) connector for Lunch Money. The idea: ask AI tools like Codex and other MCP clients questions about your budget — spending, recurring items, categories — and get answers in plain English.

Status: pre-release. The connector is not available yet, and this page does not accept tokens or signups. Not affiliated with or endorsed by Lunch Money.

Fictional preview sample data, rendered locally

How much did I spend on groceries this month?

You spent $512.40 on groceries across 9 transactions in September.

Example grocery spending by merchant
MerchantAmount
Corner Market$188.20
Green Grocer$141.75
Harbor Foods$182.45

Fictional answer based on the sample data shown here.

What recurring charges are coming up?

Three items renew in the next week:

Example upcoming recurring charges
ItemAmountRenews
Streaming plan$15.99Oct 1
Cloud storage$9.99Oct 3
Gym membership$42.00Oct 6

Fictional answer based on the sample data shown here.

Am I on track with my budget this month?

Two of three categories are under budget so far.

Example budget progress by category
CategorySpentBudget
Groceries$512$600
Dining out$268$250
Transport$96$150

Fictional answer based on the sample data shown here.

Example only — nothing is connected to a real account.

How the connection is intended to work

The connector is not built yet. This is the planned flow, in plain terms — each step is a goal, not a shipped feature.

  1. You bring your own Lunch Money token planned

    You would create a personal access token inside your own Lunch Money account and enter it in a separate browser connection flow managed by Nango. The planned connector would use the existing sign-in service at auth.n3wth.com. Never paste a token into an AI chat.

  2. Your AI client connects over MCP planned

    You would point an MCP-compatible client — for example Codex or another tool that speaks the Model Context Protocol — at the connector's endpoint:

    https://mcp.lunchmoney.sh

    Planned address only. It is not live, and there is nothing to connect to yet.

  3. Questions become read-only API calls planned

    When you ask something — "what did I spend on dining out last month?" — the connector would fetch the relevant transactions, categories, or budgets from Lunch Money's official API.

  4. Your AI client writes the answer

    The financial results would be returned to your AI client and sent to the model provider behind it to produce the answer. That means your data leaves the connector and is processed by whichever AI provider you use — see Security & privacy.

Security & privacy

Here is what this website does today and what we plan to verify before opening the connector to users.

This site, today

  • A static page with no backend of its own
  • No signup, no token entry, no forms
  • No analytics or third-party tracking added to the page
  • The chat preview runs locally in your browser with fictional data

Vercel hosts this site and may retain standard request logs, including IP addresses. No financial account is connected here.

The connector, planned

  • Sign-in through the existing auth.n3wth.com identity service
  • Browser token entry and encrypted credential storage through Nango
  • Ownership checks in the MCP server on every account request
  • An initial read-only tool allowlist enforced by the server

These are design goals, not shipped controls. Authentication, credential storage, tenant isolation, and read-only enforcement are not implemented or verified yet.

Questions

Is this an official Lunch Money product?

No. lunchmoney.sh is an independent, unofficial project. It is not affiliated with, endorsed by, or sponsored by Lunch Money. The official product is at lunchmoney.app.

Can I connect my account today?

No. The connector is in pre-release and is not available. This page has no token field, no install button, and nothing to sign up for.

What is MCP?

The Model Context Protocol is an open standard that lets AI tools call external services in a consistent way. An MCP connector for Lunch Money would let a compatible AI client ask your budget questions on your behalf.

Where would my financial data go?

Under the planned design, the connector would fetch your data from Lunch Money's official API and return it to your AI client, which sends it to its model provider to generate an answer. Your financial data would therefore be processed by the AI provider you choose — review that provider's data policies before using any tool like this.

Would it be able to change my budget or move money?

The first beta is planned for reads only. Your Lunch Money token may carry broader permissions, so the MCP server must enforce its own limits. Writes may be added later with separate scopes and approval controls. Moving money is outside this project's scope.

How would I disconnect or revoke access?

The planned disconnect flow would block further requests and delete the credential stored for this connector. That does not revoke the original token: invalidate it in Lunch Money itself to prevent it being used anywhere else.

What will it cost?

Undecided. The project is pre-release and pricing hasn't been determined.