My AI Agent LabBack to the setup guide →

Install Python and verify actual execution

Prepare Python to use this lab for real research. If a working Python 3 is already available, check it before reinstalling. Installing Python alone does not prepare every analysis tool.

Actual Windows installation video

About 2 min 11 sec · Zoomed selections and input with Korean captions. Installation waiting time is shown at 6× speed.

Original video retained: interface text and captions are in Korean. Follow the English instructions on this page; folder names and commands are unchanged.

Edited from an actual Windows recording. After responding to installation questions, it checks Python 3.14.7 and PYTHON_OK output. It does not show pip or execution inside the AI app; check those separately using the guide below.

Download the edited video (.mp4)

Windows

1. https://www.python.org/downloads/ : download and install the official Python Install Manager.

2. Open PowerShell from the Start menu and enter python --version. If only the manager is installed, Python installation may follow and requires an internet connection.

3. Python 3.x.xshould appear. Next, enter python -c "print('PYTHON_OK')"and check for PYTHON_OK.

4. Run python -m pip --version to check package-installation support. If only py works, replace python with py in these commands and continue using that name.

Official guide: https://docs.python.org/3/using/windows.html

Actual Mac installation video

About 1 min 43 sec · Zoomed buttons and input with Korean captions. Open Spotlight with Command(⌘) + Space.

Original video retained: interface text and captions are in Korean. Follow the English instructions on this page; folder names and commands are unchanged.

The edited recording checks Python 3.14.7 and PYTHON_OK and holds the final result for four extra seconds. A missing-certifi message appears; completed certificate installation is not demonstrated. Separately follow the certificate setup, pip check and in-app execution checks below.

Download the edited Mac video (.mp4)

Mac

1. https://www.python.org/downloads/macos/ : download and open the stable Python 3 macOS installer (.pkg). Do not choose a prerelease.

2. Follow the installer. After completion, run Install Certificates.command in Applications → Python 3.x. This prepares certificates used for internet access and requires a connection.

3. Press Command (⌘) + Space, search for Terminal, then run python3 --version followed by python3 -c "print('PYTHON_OK')".

4. Also run python3 -m pip --version. Do not delete or overwrite the Python managed by macOS.

Official guide: https://docs.python.org/3/using/mac.html

Complete the check inside the Claude app too

After installation, fully quit the Claude appand reopen it. In Code view,select the research-team folderand send the check request below.

About 53 sec · Actual Mac Claude app recording. Zoomed request and results with Korean captions.

Original video retained: interface text and captions are in Korean. Follow the English instructions on this page; folder names and commands are unchanged.

The video starts with research-team already selected in Code view. Claude reports Python 3.12.2, a calculation result and pip. This differs from Python 3.14.7 in the installation video, so check the executable location actually used. The analysis-tool list is specific to the recorded computer and is not a default bundle for every user. Detailed individual execution records are not expanded in the video.

Download the Claude app check video (.mp4)

Request to send to Claude

Run Python 3 in this app and report its version and executable location. Verify a simple calculation and check that pip runs with the same Python. Do not install anything or analyze data yet; report the actual results.

The AI uses the command that works in this environment to run -c "import sys; print(sys.version); print(sys.executable); print(1+1)"and -m pip --version. Verify an actual calculation result of 2 and the executable location, not just a version string. Success in a user's terminal does not establish readiness if execution fails in the app. Use the same verified executable for later setup. Quote the full path appropriately if it contains spaces.

This check does not require additional installation, network lookups or research data access. However, Windows Install Manager may attempt a download on first use if Python itself is missing. If installation begins, tell the user rather than concealing it.

If using the Codex app, open the team folder there and request the same check.

Troubleshooting

- Command not found or only Microsoft Store opens: Python execution has not been verified. Check that official installation finished and reopen PowerShell, Terminal and the AI app. On Windows, consult Troubleshooting in the official documentation.

- Terminal succeeds but the app fails: Compare the working Python executable path with the app's path. Another computer or an isolated environment may need its own setup. Avoid repeatedly reinstalling.

- pip is missing: Report Python execution separately from package-installation readiness. Check the official installation and the interpreter actually selected. Do not mark all functionality ready.

- Internet certificate errors on Mac: Check completion of Install Certificates.command and the Python location in use. Do not disable certificate verification to resolve the error.

- No installation permission: Ask your computer administrator to install it. Record the failure and do not automatically mark setup complete.

Additional preparation when research begins

Create a separate analysis environment per project so tools from different studies do not interfere. Have the AI prepare only tools needed for the requested analysis and record Python and package versions. This is separate from project setup. R-based analysis also requires R. Do not install every MedSci tool or audit the entire library at once.

Verification scope: This guidance is based on official installation documentation. The complete path from installation on a fresh Windows or Mac computer to execution inside the app still requires separate testing.