Lunch Money MCP (Unofficial)
io.github.n3wthPublished by GitHub account n3wthv0.1.0Updated Oct 27, 2026
Read-only Lunch Money accounts, transactions, categories and budgets. Unofficial connector.
Overview
Unofficial hosted MCP connector that lets an AI assistant read your Lunch Money accounts, transactions, categories, budgets and recurring expenses.
- What you can do
- Read accounts and balances, keeping each currency separate
- Query transactions, including by date range and grouped by category
- Review categories and compare spending against budgets
- List recurring expenses
- Connect, replace or remove the Lunch Money connection from the assistant
- When to use it
- Useful when you want to ask an assistant conversational questions about your own Lunch Money finances, such as last week's spending by category, budget comparisons, or recurring charges, without exporting data by hand.
- Requirements
- A Lunch Money account and a personal access token created on Lunch Money's Developers page. An MCP client that supports remote servers with OAuth sign-in; you sign in at the connector's auth site and enter the token in the provider's browser connection page. The hosted endpoint is in beta.
Installation
In SourceWeft
- Open Lunch Money MCP (Unofficial) in the dashboard and add it to a workspace.
- Connect the required credentials. They are stored privately and never shown on public pages.
- Enable the server for the chats that should use its tools.
Web executable via Streamable HTTP. Remote servers run from the web runtime once configured in a workspace.
Use with an agent
Paste this prompt into Claude Code, Codex, Cursor or another agent — or into SourceWeft chat. The agent reads this server's install guide, shows you its publisher, endpoint and sign-in, and adds it to your client once you agree. Keys never go into a command.
Read https://sourceweft.com/mcp/io.github.n3wth%2Flunchmoney-mcp/install.md and connect the MCP server it describes to my client. Before changing any config, show me its publisher, endpoint and sign-in method and wait for my OK. Never put an API key in a command or in a file I would commit; ask me where it should come from.Other MCP clients
Add this to your client's mcpServers config. The client completes OAuth sign-in on first connect.
{
"mcpServers": {
"lunch-money-unofficial": {
"type": "http",
"url": "https://mcp.lunchmoney.sh/mcp"
}
}
}README
Lunch Money for Agents
[Lunch Money for Agents icon]Ask your AI agent about your Lunch Money data.
Website · Setup guide · Privacy · Report an issue
I built this to make Lunch Money easier to use with AI agents. It connects your account to tools that support Model Context Protocol (MCP), so you can ask about your spending, budgets, accounts, and recurring expenses in conversation.
This is an unofficial, open-source MCP integration by n3wth. It is not affiliated with, endorsed by, or sponsored by Lunch Money.
What you can ask
- “Show my transactions from last week, grouped by category.”
- “How does this month's spending compare with my budget?”
- “List my recurring expenses.”
- “What are my account balances? Keep each currency separate.”
The integration gives your agent access to the underlying records; your agent uses those results to answer. Financial access is read-only: it cannot edit transactions, change budgets, or move money. Connection tools can link, replace, or remove your connection to the service.
Connect your account
You'll need a Lunch Money account, a personal access token, and an MCP client that supports remote servers with OAuth sign-in.
- Add the server using your client's setup instructions below.
- Sign in at
auth.n3wth.com. This creates your connector session, separate from your Lunch Money login. - Ask your agent to connect Lunch Money. It should run
lunchmoney_connectand give you a browser link. - Open the link and enter your token in Nango's connection page. Create the token on Lunch Money’s Developers page by clicking Request New Access Token (label and usage are optional). Never paste it into chat or your MCP configuration.
- Try a small request, such as “List my Lunch Money accounts.”
The hosted endpoint is:
Client-specific installation
Choose your app on the website, or install directly:
Claude Code — run inside Claude Code:
Codex — run in a terminal:
For the Codex app, use Plugins → Create menu → Add marketplace with
n3wth/lunchmoney-mcp, then install Lunch Money.
Cursor — Add to Cursor,
the Cursor directory listing,
or select Cursor on the website.
Start a new session after installing, sign in, then ask “Connect Lunch Money”.
Custom marketplaces are separate from curated directory approvals. Full instructions and manual alternatives:
The hosted service is in beta. Direct setup is available; a directory submission does not mean a plugin is approved or listed. See the submission record for review status and the runbook for outstanding release checks.
Where your data goes
Signing in and connecting Lunch Money are separate steps:
The server retrieves your token from Nango when needed and enforces financial reads only. Your token may have broader permissions in Lunch Money itself. Financial responses are not saved in the connector's database, but your AI client's privacy and retention settings still apply to the results it receives.
To disconnect, ask your agent to run lunchmoney_disconnect. If deletion fails,
retry until it succeeds. Revoke the original token in Lunch Money if you also
want to prevent its use elsewhere. Removing a client plugin does neither.
Read the privacy notice for storage, retention, and provider details, and the service terms before connecting your account.
A couple of things to watch
Transactions arrive in pages. One response may not cover the whole date
range. The agent needs to follow next_offset until it is null before
presenting a complete total.
Currencies stay separate. Adding USD and CAD amounts together does not produce a useful total. Ask for separate totals unless you explicitly want a conversion.
AI-generated answers can be wrong. Check the source records in Lunch Money when an answer matters. The user guide lists the tools and explains connection troubleshooting.
Work on the project
Use Node.js 22 or later. The site is static HTML and CSS with a small React conversation preview built from Vercel AI Elements.
This serves the site at http://localhost:4173. To build it without starting
a server, run npm run build; the output goes into dist/.
Run the server checks in dependency order:
Staging and production use separate configurations and databases. Default Worker commands target staging. Follow the runbook for deployment, credentials, migrations, and rollback.
Context7 documentation indexing
context7.json configures which documentation Context7 indexes. It includes this README, the documentation overview, connection, tools, and privacy guides, and the Codex, Claude Code, and Cursor setup guides. Internal plans, operational records, and agent instructions are excluded. Context7 always includes root Markdown files unless excluded by name.
After merging this file into the repository's default branch, open the
Context7 library settings,
switch the configuration source to the repository's context7.json if it is
currently managed in the dashboard, and refresh the library. Verify the indexed
sources include the setup guides and omit the excluded records.
See Context7's configuration reference
for supported fields. This controls indexing only; a hosted Docs7 site uses a
separate docs.json configuration.
Contribute
Bug reports, clearer setup instructions, and compatibility fixes are welcome. For a bug, include the client you used, what you expected, and a minimal example with sample data. Leave tokens and personal financial records out of issues.
Open an issue before a larger change so we can agree on the scope. For a pull request, explain the problem and how you checked the fix. Keeping financial access read-only is a core constraint of this project.
Source: README.md at commit 1676593
Tools
0Version history
1- v0.1.0LatestSep 16, 2026
Score
85% of the applicable checks met
How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.
- Reached in the last 30 daysRequired25 pts
- Auth declaredRequired10 pts
- Tools listed15 pts
Answer tools/list without a credential, or ship a package that starts in a sandbox without one.
- DescriptionRequired15 pts
- Logo10 pts
- Documentation15 pts
- License10 pts