org.lunenburgbudgetproject/archive
The Lunenburg, Massachusetts town and school budget, every figure traced to its source document.
Install
Add this to your client's MCP configuration (Claude Desktop: claude_desktop_config.json; Cursor: .cursor/mcp.json; VS Code and others use the same "mcpServers" shape).
claude mcp add --transport http archive 'https://lunenburgbudgetproject.org/mcp'{
"mcpServers": {
"archive": {
"type": "http",
"url": "https://lunenburgbudgetproject.org/mcp"
}
}
}Documentation
From the project's README on GitHub, refreshed weekly. Links open on GitHub.
Contents
The Lunenburg Budget Project
An independent, checkable archive of the Lunenburg, Massachusetts town and school budget: 43,619 documents, 292 datasets, and 12,216 agendas and sets of minutes across 51 town boards. Not affiliated with the Town of Lunenburg, the School Committee or the school district.
The site is https://lunenburgbudgetproject.org. This repository is everything behind it.
Start there: the data API · an MCP server · every published address · the guide for agents.
If you are an agent, start here
Read fy28/public/llms.txt first. It is written for you:
what this archive holds, what it does not, and the specific ways to get a confident wrong
answer out of the data.
If you can reach the site, use it — every address below has a live equivalent and the site serves smaller, indexed forms. If you can only reach GitHub, fetch raw files directly; no clone is needed:
https://raw.githubusercontent.com/tjloughlin16/lunenburgbudgetapp/main/
| you want | fetch |
|---|---|
| what this holds, and its limits | fy28/public/llms.txt |
| every dataset, with row counts and sizes | fy28/public/api/tables.json — 118 of them |
| the grain of every table, and the traps | fy28/public/api/schema.json |
| every figure the site computes | fy28/public/data/model/index.json — 38 sections, one file each |
| every source document, with its sha256 | sources/data/archive-manifest.csv |
| which meeting documents mention a word | fy28/public/minutes/find/README.txt |
Nothing in fy28/public/api/ is larger than one fetch. Anything that would be is split
— staff_roster_entries.json is an index and staff_roster_entries/2022.json is that year.
The documents themselves are not in this repository. 43,619 files, 1.47 GB, live
in a public object store; sources/data/archive-manifest.csv lists every one with its
sha256 and python3 scripts/sync_archive.py --pull fetches them. The extracted text IS
here, under sources/*/text/, and that is what the analysis reads.
Where things are
sources/ | the archive, keyed by how a document reached us — see sources/README.txt |
sources/data/ | 292 datasets extracted from those documents, CSV |
sources/analyses/ | the written analyses, each with a verifier script |
model/ | the projection — python3 model/export.py writes fy28/src/data/model.json |
scripts/ | extraction, verification and publishing |
fy28/ | the site: React app, Cloudflare Pages Functions, and everything published |
notes/ | working notes; notes/reference/ is generated |
plans/ | what is being built and why |
The rules this project works under
CLAUDE.md carries them in full and each one exists because it was learned by getting it
wrong here. The four that matter most to anybody reading the data:
- Never mix budgets with actuals in one calculation. They differ by up to 59% on some lines, and a growth rate measured from one to the other is partly growth and partly the step between them.
- A budget line is NET, and it is dollars. It is what the town must raise after grants, fees and state aid — not what the thing costs, and never a count of people.
- Only facts are stated as facts. A number computed from the data is a fact; an explanation for why it moved is a hypothesis, and it is labelled as one.
- Every figure is interpolated from the model, never typed into prose. Including the counts in this file.
Running it
npm install --prefix fy28 python3 scripts/sync_archive.py --pull # the documents; the manifest says what python3 model/export.py # -> fy28/src/data/model.json npm run --prefix fy28 build:site # -> fy28/dist
CLAUDE.md lists every check. The ones that guard what this file describes:
python3 scripts/build_readme.py --check # this file, against the repository python3 scripts/build_api.py # every dataset, fetchable node fy28/scripts/check-agents.mjs # every advertised URL answers, and can be read python3 scripts/check_archive_storage.py # the archive against its manifest
Generated by scripts/build_readme.py. Edit that, not this.
Frequently asked questions
What is org.lunenburgbudgetproject/archive?
The Lunenburg, Massachusetts town and school budget, every figure traced to its source document.
How do I install org.lunenburgbudgetproject/archive?
It runs as a remote server. In Claude Code: claude mcp add --transport http archive 'https://lunenburgbudgetproject.org/mcp'. For other clients, paste the JSON configuration from this page into their MCP settings.
Does org.lunenburgbudgetproject/archive need an API key?
Its registry entry lists no required key or secret. Some features may still need an account with the service it connects to.
Is org.lunenburgbudgetproject/archive open source?
Its code is public on GitHub, but the repository does not declare a standard license.
Who maintains org.lunenburgbudgetproject/archive?
It is published by tjloughlin16. AI.tn lists it but does not maintain or endorse it.
From the official MCP Registry. Listing is not an endorsement: review a server's code and permissions before giving it access to your data.