Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@voiceledger-mcp What did I sell today?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
A self-hosted MCP server that gives an assistant a cash trader's books.
Ask it what you sold today, what is running out, and who still owes you, and it answers from the same append-only ledger the trader actually writes into. Built for the Alexa+ track of Build, Ship, Shape: Amazon Developer Hackathon.
Most of the shops in a Nigerian market run on memory and a paper book. The book is honest and the memory is not, and neither of them can answer a question while the owner's hands are full of rice. The gap is not a dashboard. It is being able to ask out loud and get a number you can stand behind.
Related MCP server: manager-mcp
Five tools and one resource, over MCP:
record_entry writes a sale, expense, restock, credit or repayment.
day_summary gives the day's totals and the cash that should be in the box.
summarise_day_for_speech gives the same figures, phrased to be read aloud.
low_stock lists what is running out, counted from restocks minus everything that left.
outstanding_debts lists who still owes, largest first.
ledger://entries exposes the raw book as JSON.
Money is integer kobo. No float ever touches a naira figure, so a hundred small entries do not drift.
The ledger is append-only. A mistake is corrected by writing another entry, the way a paper book works.
Goods on credit are not sales. They leave stock, they raise a debt, and they stay out of the day's takings.
A credit entry with no customer name is refused, because an unnamed debt is an unrecoverable one.
A model never phrases a number. Figures are computed here and passed through verbatim; Bedrock only warms up the sentence around them, and if it is down the deterministic line goes out unchanged.
cp .env.example .env
npm install
npm test
npm run devThe Streamable HTTP endpoint is then at http://localhost:8787/mcp, with /healthz alongside it.
Set MCP_AUTH_TOKEN to require Authorization: Bearer <token>; leave it unset to run open locally.
TypeScript and Node 20+, the Model Context Protocol SDK over Streamable HTTP (spec 2025-11-25), Express for the endpoint, Zod for tool schemas, Vitest for the tests. Amazon Bedrock is optional and used only for speech phrasing.
Built for the Alexa+ track of Build, Ship, Shape: Amazon Developer Hackathon (deadline 23 October 2026), and entered in the AWS Builder and Open Source mini challenges. MIT licensed.
This server cannot be deployed
Deterministic signed verification of numeric & financial claims for AI agents & spreadsheets.
Query your books, company data, business health, and ship invoices from your AI assistant.
Personal, family and business books over fiat and crypto. Reads and records; it cannot pay.
Personal finance ledger for AI agents — query spending, track bills, forecast cash flow.
