Gets the data out, trains or fine-tunes the model, and evaluates it before anybody calls it an improvement.
Built for: The data scientist who spends more of the week moving files than modelling, and has no evaluation harness.
The button opens the checkout, where 7 servers and 5 skills are listed one by one with what each does to the bill — free, already yours, monthly or a one-off licence. Nothing is charged until you confirm it there, in Stripe’s own card frame on that page rather than a redirect, and each paid member keeps its own budget cap.
The reason to buy a stack rather than five listings: each line below needs more than one member connected at the same time.
Load a CSV or an Excel file, profile the columns and say what is missing before any modelling starts.
Query Snowflake read-only, and read the files that never made it into the warehouse from Drive.
Fine-tune a language or vision model through Hugging Face infrastructure, with the training recipe written down.
Evaluate an agent or a prompt against a defined harness instead of a demo that happened to work.
What each part contributes, and why it was picked over the obvious alternative. Prices and permissions are read from the listings, so nothing here can disagree with the catalogue.
Loads a CSV, queries it and describes the columns — the first ten minutes of every dataset, without writing a loader.
The Hugging Face side of the desk: models, datasets and the jobs that run against them.
Read-only access to the production database for the features that only exist in the application's own tables.
Warehouse queries with key-pair or SSO authentication and no write path, which is the right shape for exploratory work.
The spreadsheets and exports colleagues send instead of loading, searched and read where they already are.
A baseline to measure against: the hosted model the fine-tune has to beat to be worth its cost.
Statistical analysis and preprocessing over Excel and CSV files, for the preparation work that happens before a notebook.
Keeps the analysis in pandas, matplotlib and seaborn the way the rest of the field writes it, so the notebook is readable by the next person.
Training and fine-tuning through TRL or Unsloth on Hugging Face Jobs, with the recipe recorded rather than remembered.
The evaluation harness for agents and LLM applications, which is the difference between an improvement and an impression.
A worked example of querying a real personal dataset — an Apple Health SQLite database — in Markdown, JSON or FHIR.
Turns an arXiv paper into a minimal citation-anchored implementation, which is how a method gets tested instead of cited.
Every member is a subscription or a licence bought once, so the monthly figure is a price rather than an estimate: what moves it is adding or dropping a member, not how hard the stack is worked. The one assumption is that a one-off licence is spread over a year so it can sit in the same column as a subscription.
| Part | What you are paying for | Monthly, as quoted |
|---|---|---|
| Csv Analytics | Free | — |
| Huggingface Client | Free | — |
| Postgres | Free | — |
| Snowflake Readonly | Free | — |
| Google Drive (official) | Free | — |
| Openai | Free | — |
| Excel Csv | Free | — |
| Skills | ||
| Data Analysis in Jupyter | $99 · $8.25/mo over 12 months | $8.25/mo |
| Hugging Face LLM Trainer | Free · context cost only | — |
| DeepEval | $39 · $3.25/mo over 12 months | $3.25/mo |
| Health Data | $19 · $1.58/mo over 12 months | $1.58/mo |
| Paper2Code | $29 · $2.42/mo over 12 months | $2.42/mo |
| Everything above | $0 of servers plus $16 of skills, the same in a quiet month and a busy one | $16/mo |
Subscriptions at their monthly plan price; one-off licences spread over 12 months. One-off purchases in this stack total $186 — Data Analysis in Jupyter $99, DeepEval $39, Health Data $19, Paper2Code $29 — paid once and spread here so they sit in the same column as a subscription. Everything arrives on one mcprush invoice, taken by Stripe from the card on your account, not one per publisher — mcprush.com is the merchant of record and each publisher is paid out of it.
Install only these and the bill is nothing: 7 servers and 1 skill.
Left out, and what goes with it:
Nothing in this stack carries a call allowance, so nothing here can run out before the month does. The bill is decided when you install it, not when you use it.
Budget caps are set per install and enforced at the gateway, so a retry loop is refused at the cap rather than left to run through an allowance overnight.
One step per part, in the order they are useful: connect what the work reads before what it writes, and install the skills that decide how the work is done last. Each step is a command you can read before you run it.
npx mcprush@latest stack add ml-deskNothing in this stack installs from one command today: 12 members are either paid, run from its own source, or a skill with its own command — the steps above name each one. Nothing is connected until you approve it.
Installed together, these tool surfaces add up. It is the first thing a security reviewer asks for, so what has been counted — and what nobody has counted yet — is on the page rather than in a PDF.
| Member | Tool surface | Write tools |
|---|---|---|
| Csv Analytics | not imported | not recorded |
| Excel Csv | not imported | not recorded |
| Google Drive (official) | not imported | not recorded |
| Huggingface Client | not imported | not recorded |
| Openai | not imported | not recorded |
| Postgres | not imported | not recorded |
| Snowflake Readonly | not imported | not recorded |
A stack's blast radius is the union of its members, not the worst of them. That union cannot be taken here, because 7 members of 7 have no imported tool surface — so the figure a review asks for is missing rather than low, and a member marked not imported is one nobody has counted rather than one that cannot write.
A stack is a default, not a verdict. These are the substitutions the maintainer would make, and what each one costs or saves.
For a data lake on S3 rather than a warehouse: the same read-only querying, billed per scan instead of per credit.
Where the files arrive as spreadsheets rather than as clean CSVs, the Excel reader handles them with pagination and the CSV tooling becomes redundant.
Written by the maintainer, kept on the page rather than in a support thread.
It cannot deploy a model. Training and evaluation are here; serving is an infrastructure decision this stack does not make.
The warehouse connection is read-only. A feature table has to be built by whoever owns the pipeline.
It will not tell you a model is good. The evaluation gives numbers against a harness you defined; the threshold is yours.
A stack has an owner: whoever keeps it re-checks the combination when a member changes, and the members themselves are published by the people named on each row.
Designs the endpoint, wires the auth and the payment, and reviews the shape before it is public and permanent.
Turns a bug report into a reproduction that fails, before anybody writes a line of the fix.
Writes the migration, proves it on a branch, and says what it will do to the table before it touches the real one.