How to Install the Data Scientist Claude Plugin (Cowork & Code)

Don't want to install anything? You don't have to. The free web tools for data scientists on The AI Career Lab give you the same pre-built workflows in your browser — five runs a day on a free account, no terminal, no install. Create a free account if you'd rather skip straight to the tools.

The Data Scientist Claude plugin from The AI Career Lab is the fastest way to turn Claude into a DS-aware co-worker. It bundles 4 pre-built slash commands and 2 auto-activating skills that already know what working data scientists actually do — A/B test design with power analysis, dataset profiling with leakage detection, Model Cards aligned to the Mitchell et al. 2019 framework, and stakeholder briefs that don't bury the lede.

The plugin works in both runtimes Claude offers:

• Claude Cowork — the chat product at claude.ai. The easiest path. No terminal, no install of anything else. Requires a paid Claude plan (Pro, Max, Team, or Enterprise).
• Claude Code — the terminal CLI. Particularly useful for data scientists since Claude Code can read your notebooks and Python files directly. Requires Node.js + npm. Free or paid Claude account both work.

What this plugin gives you

A normal Claude session is a blank chat window. A plugin replaces that with slash commands — pre-built shortcuts that contain the structure, formatting rules, and DS conventions for a specific task.

In this plugin you get:

• /experiment-design — computes sample size, power, MDE, and duration for an A/B test with formula shown, pre-committed analysis plan, and SRM checks
• /dataset-profile — generates runnable profiling checks for missing values, outliers, class imbalance, and the four major leakage patterns (target, train-test, temporal, proxy)
• /model-card — produces a Model Card in the Mitchell et al. 2019 format with fairness audit plan and reviewer questions
• /stakeholder-brief — translates findings into a brief tailored to the audience (exec, product team, customer success, legal) with headline + ask up front

Which runtime should you use?

Use Claude Cowork if:

• You already use claude.ai for chat work
• You prefer a graphical interface to a terminal
• You don't want to install anything else on your machine
• You have a paid Claude plan (Pro, Max, Team, or Enterprise)

Use Claude Code if:

• You're already working in a terminal alongside your notebooks
• You want Claude to read your actual Python files, notebooks, or SQL directly when designing experiments or profiling data
• You're on a free Claude account (Claude Code plugin install does not require a paid plan)

For data scientists specifically, Claude Code has a strong advantage: it can read your repo. You can run /dataset-profile and have Claude read the actual schema from your data dictionary file rather than re-paste it.

Path A: Install in Claude Cowork (easiest, no command lines)

1. Open claude.ai in your browser and sign in.
2. Make sure you're on a paid plan (Pro, Max, Team, or Enterprise). The plugin marketplace is gated to paid plans.
3. Open the plugin settings and click Add marketplace.
4. In the modal, paste this into the URL field:

alexclowe/awesome-claude-cowork-plugins

5. Click Sync. Claude Cowork pulls the marketplace catalog from the public alexclowe/awesome-claude-cowork-plugins repo.
6. Browse the marketplace. Find the Data Scientist plugin and click Install.

Path B: Install in Claude Code (terminal CLI, two commands)

One-time Claude Code setup

1. Install Claude Code. Open a terminal and run:

npm install -g @anthropic-ai/claude-code

If you don't have npm, install Node.js first (download the LTS version, run the installer).

2. Sign in once. Run claude in your terminal and follow the sign-in flow.

Install the plugin

Inside Claude Code, run these two commands:

/plugin marketplace add alexclowe/awesome-claude-cowork-plugins
/plugin install data-scientist@awesome-claude-cowork-plugins

To verify, type /help and you should see the new commands listed alongside Claude Code's built-ins.

Your first command

For most data scientists, the highest-leverage starting point is the experiment design command — every team has an A/B test backlog:

/experiment-design

When you run it, Claude asks for your hypothesis, primary metric, baseline rate, MDE, and traffic available. The output is a power analysis with the formula shown (so you can verify the math), a pre-committed analysis plan, and the specific SRM and balance checks to run.

Commit the analysis plan to your team's experiment tracking system before launching the test. That's the discipline that prevents "directional results" decisions on underpowered tests.

First 4 workflows to try after install

1. A/B test design — /experiment-design in the plugin, or /tools/data-scientist-experiment-design on the web. Run this for the next experiment in your backlog
2. Dataset profiling — /dataset-profile in the plugin, or /tools/data-scientist-dataset-profile on the web. Run this on the next fresh dataset before training anything
3. Model Card — /model-card in the plugin, or /tools/data-scientist-model-card on the web. Run this for any deployable model before model risk review
4. Stakeholder brief — /stakeholder-brief in the plugin, or /tools/data-scientist-stakeholder-brief on the web. Run this for any cross-functional readout

Skills you get for free

This plugin also includes Claude Skills that activate automatically when you're working on relevant tasks:

• Verifiable Math Discipline — when generating statistical computations (power analysis, confidence intervals, sample size), shows the formula or approach rather than just the final number. So the DS can verify the math
• Uncertainty Framing — when summarizing findings for stakeholders, automatically structures output with the uncertainty named directly rather than buried at the end

The first time the verifiable-math skill auto-activates and shows you the variance formula it used for an unequal-variance two-sample test, you'll see the difference between using Claude generically and using Claude as a DS co-worker.

Where to go from here

• Claude Cowork playbook for data scientists — the full setup guide with the five high-leverage prompt structures
• Free web tools for data scientists — same workflows, browser-based, no install required
• Best AI Tools for Data Scientists in 2026 — the broader landscape including Claude, notebook environments, experimentation platforms, eval tools
• AI for Data Scientists guide — the working playbook for offloading the tedium while owning the judgment

Common issues

I don't see the "Add marketplace" option in Claude Cowork. You need to be on a paid plan. If you're on a free plan, install via Claude Code instead — it's free.

The marketplace sync failed in Claude Cowork. Double-check the URL field has exactly alexclowe/awesome-claude-cowork-plugins — a GitHub owner/repo format, no https://, no trailing slash.

The slash commands aren't showing up after install. Start a new chat (Cowork) or type /help (Claude Code). If you see the built-in commands but not the new ones, the install didn't complete — re-run the install.

I want to use this without installing anything. Use the free web tools instead. Same workflows, browser-based, no install or paid plan.

Ready to try it without installing

Create your free AI Career Lab account and try the data scientist tools today. No credit card required.

---

This article is general guidance for data scientists. Statistical methodology, fairness frameworks (Mitchell et al. 2019 Model Cards, Hardt et al. 2016 Equal Opportunity), and regulatory frameworks (EU AI Act, NIST AI RMF, US state AI laws) are evolving. The original methodological papers and your organization's data science standards remain authoritative references. The Model Card tool produces one input to a regulatory review; it does not constitute compliance certification.