LevelUpWith

Explore / Claude Science: A Researcher's Working Guide / Lesson 2

Lesson 2: Installing and Running a First Analysis

Install the app on your operating system, complete sign-in and the setup wizard, and run one small analysis from prompt to saved result.

Free to read, no account needed. Part of Claude Science: A Researcher's Working Guide.

By the end of this level you'll be able to

  • Install and open Claude Science on macOS, Windows or Linux, noting the Linux sandbox dependencies and the main differences on Windows
  • Complete the setup wizard choices for connectors and skills, website access and memory, knowing each can be changed later in Settings
  • Run a first analysis on the Example project or a small non-life-science dataset (for example, a regression on regional economic data) and confirm the result appears as an artifact in Files

What you will do

In this level you install Claude Science, sign in, answer the setup wizard, and run one small regression from prompt to saved file. Allow about 30 minutes, most of it waiting for the first launch.

The app is in beta, so labels and steps can change. If your screen differs from this lesson, follow Anthropic's current Get started page.

Check the requirements

Confirm these before you download:

  • Plan: Pro, Max, Team or Enterprise. The Free plan cannot sign in. On Team and Enterprise, an Owner must turn Claude Science on for the organisation first.
  • Operating system: macOS 13 or later (Apple silicon or Intel), Windows 11 (x64), or x64 Linux on a glibc-based distribution.
  • Disk: about 5 GB free for the runtime and starter environments.
  • Region: a country where Anthropic offers claude.ai.

Install on your operating system

Claude Science is a local application, not a website. You always open it from your own computer.

macOS

  1. Download the installer from claude.com/product/claude-science and double-click it.
  2. Wait a few minutes while the first launch sets up the runtime and the starter Python and R environments.
  3. The app opens as a new tab in your default browser. If no tab appears, click Claude Science in the Dock.

If macOS says the application is not supported, you have the build for the wrong processor. Go back to the download page and choose Intel or Apple Silicon to match your Mac.

Windows

  1. Download the installer from the same page and open it. It installs for your user account and adds Claude Science to the Start menu and desktop.
  2. Wait for the first launch. It sets up the sandbox and downloads a window engine of about 150 MB, so the first start is slower than later ones.
  3. The app opens in its own window, not a browser tab. The starter environments keep setting up in the background for several minutes.

The main differences on Windows:

  • Shell: Claude runs shell commands with PowerShell.
  • One-time Windows permission: the app may ask for it so that cells can run git and Command Prompt scripts. Python and R cells work whether you answer Yes or No, and you can grant it later under Settings > Permissions.
  • Folders: only folders on local NTFS or ReFS drives can be shared with Claude. Network shares, mapped drives, FAT or exFAT USB sticks and a whole drive such as D:\ cannot. Copy such files to a local folder first.
  • Closing the window: the app keeps running, with an icon in the taskbar notification area. Use Quit Claude Science from that icon to stop it.

The Windows app does not need WSL. Running the Linux version under WSL 2 is a separate, optional route.

Linux

The sandbox that isolates running code needs two system packages: bubblewrap 0.8.0 or later and socat. Installing them needs sudo.

  1. Install the dependencies. On Ubuntu or Debian:
sudo apt-get update && sudo apt-get install -y curl bubblewrap socat

On Fedora or RHEL use dnf, and on Arch use pacman, with the same three package names.

  1. Run bwrap --version and confirm it reports 0.8.0 or later. Ubuntu 24.04's repositories carry a new enough version. Ubuntu 22.04's do not.
  2. Run the installer:
curl -fsSL https://claude.ai/install-claude-science.sh | bash
  1. Start the app:
claude-science serve
  1. Open the local URL it prints in a browser on the same machine. If you lose it, claude-science url prints a fresh one.

An error that starts with Sandbox unavailable points to bubblewrap: it is missing, too old, or blocked from creating user namespaces.

Sign in

  1. Sign in with your Claude account. No API key is needed.
  2. On Windows, click Sign in on the web, confirm with Continue, and approve the sign-in in your browser.
  3. If you approve in the browser and the app does not continue, select Paste code instead, then Open on web. Copy the whole code and paste it into the app.

Complete the setup wizard

After sign-in, the wizard asks for three choices:

  • Connectors and skills: which external data sources and written method instructions are enabled.
  • Website access: which websites Claude may reach.
  • Memory: whether memory is on.

None of these is final. All three can be changed at any time in Settings, so accept what looks reasonable and move on. Later levels cover each one in depth.

Run your first analysis

You can open the Example project or create a new project. This worked case uses a new project and a non-life-science dataset.

  1. Prepare a small CSV in a local folder. Example: regions.csv with one row per region and columns region, gdp_per_capita, unemployment_rate and tertiary_education_share. Twenty to fifty rows is enough.

  2. Create a new project and start a conversation.

  3. Point Claude at the folder by typing its path, or use the @ picker in the message box.

  4. Type a specific request, for example:

    Using regions.csv in that folder, fit an ordinary least squares regression of gdp_per_capita on unemployment_rate and tertiary_education_share. Save a coefficient table and a residuals-versus-fitted plot.

  5. A folder-access card appears. Read it and choose whether to allow it.

  6. A code-execution card appears when Claude proposes running code. Read it and choose whether to allow it.

  7. Wait for the run to finish. If the starter environments are still setting up, the first run may be slow.

Treat the two cards as plain yes-or-no decisions for now.

Check the record

  1. Open Files in the sidebar. It shows a searchable grid of the project's artifacts.
  2. Confirm that the coefficient table and the plot are listed. An artifact is a file Claude saved into the project, stored on your computer.
  3. Click the plot to open it and check that the axes match your request.
  4. If an output is missing, ask Claude to save it. Other files written during a session are temporary and are cleared a few hours after the session ends.

A result that appears only in the chat text is not yet a saved result. The check is that it is listed in Files.

Key takeaways

  • macOS and Linux open the app in a browser tab. Windows opens it in its own window.
  • Linux needs bubblewrap 0.8.0 or later and socat before claude-science serve will start.
  • On Windows, only folders on local NTFS or ReFS drives can be shared, and shell commands run in PowerShell.
  • The wizard's choices for connectors and skills, website access and memory can all be changed later in Settings.
  • A first analysis is complete when its outputs appear as artifacts in Files.

A 6-question knowledge check follows this lesson in the app, free and without an account.

Keep going: levels 3 and 4 are free

Create a free account to open the next two levels, save your progress and pick up where you left off. No card needed.

Next: Level 3, Projects, Sessions, Workspace and Data Directory.

Create a free account →See the whole course

More like this

  • Production Agents on Claude: Design, Build, Run

    Design, build and operate Claude agents that hold up in production: choose between workflows and agents, use the Claude Agent SDK and Managed Agents well, and…

    Course · Claude · AI & Agents · Expert · 64 levels · 4 free · Professionally curated by levelupwith.com

  • Finding and Fixing Vulnerabilities with Claude

    Scan a practice repository with Claude's security layers, verify and triage what they report, and land patches safely on a branch. You will leave with a repeat…

    Course · Claude · IT Support, Networking & Security · Capable · 18 levels · 4 free · Professionally curated by levelupwith.com

  • Claude at Work: No-Code Skills for Knowledge Workers

    Learn to brief Claude clearly, check what it gives you, and set up Projects, Artifacts, Skills, memory, connectors and Research for real work in sales, marketi…

    Course · Claude · AI & Agents · Capable · 18 levels · 4 free · Professionally curated by levelupwith.com

  • Delegating Real Work to Claude: Tasks, Files and Apps

    Learn to hand Claude multi-step work on your real files, browser and Office apps, with clear briefs, firm limits and approval gates, then review the results so…

    Course · Claude · AI & Agents · Capable · 18 levels · 4 free · Professionally curated by levelupwith.com

  • Prompting Claude: Instructions, Context and Testing

    Write prompts that Claude follows reliably, manage what goes into its context window, and test and improve your prompts with evidence instead of guesswork. Eac…

    Course · Claude · AI & Agents · Capable · 20 levels · 4 free · Professionally curated by levelupwith.com

  • Finished Work with Claude Artifacts, No Code

    Brief, refine, brand, check, share and export finished designs, decks, documents, dashboards and motion pieces in Claude without writing code. Every level ends…

    Course · Claude · AI & Agents · Capable · 18 levels · 4 free · Professionally curated by levelupwith.com