First Project Tutorial: Explore, Then Build

Who this is for: Workshop participants who have completed D.tutorial.setup and have Claude Code running.

What you will do:

Time: 50–60 minutes total (Part A: 20 min, Part B: 35 min)

Part A: Clone and explore the workshop repository

The workshop uses a repository of documents about AI-assisted research. You will clone it, open it in Claude Code, and ask Claude questions about it — your first real session with a multi-document project.

A1. Clone the repository

Open PowerShell and navigate to your Documents folder:

cd ~/Documents

Clone the workshop repository (you need GitHub team access for this — your presenter has added you):

git clone https://github.com/[ORG]/claude-lecture.git

Move into the folder:

cd claude-lecture

You now have a local copy of all the workshop materials on your machine.

A2. Open Claude Code in the project

claude

Claude Code reads CLAUDE.md automatically — the project configuration file that tells Claude what this project is and how to work with it. Every new session in this folder starts from that context.

You are now in a live session with a project that has over 60 documents.

A3. Ask Claude about the project

Try these prompts in order (copy them exactly, or adapt them):

Orientation:

Read CLAUDE.md. In 2 sentences: what is this project and who is it for?

Document exploration:

Explain B.adoption-spectrum.md. What are the five levels?
Give me one concrete example of what someone can do at Level 2 that they cannot do at Level 1.

Personalised recommendation:

I am a historian who doesn't code and I have 2 hours to read.
Which 3 documents from this project should I read first, and why?

Cross-document question:

What are the three biggest risks of using AI in research, according to this project?
Which documents cover these risks in the most depth?

A4. Your own question

Now ask something you actually want to know. Look at the document list (type dir to see files, or ask Claude to list them) and pick something that interests you.

Some ideas:

Type /exit when you are done exploring.

Part B: Your first research project

Now you will create a project folder for your own research. This is where the practical value of Claude Code begins: not working with someone else's documents, but with your own.

B1. Create the project folder

Navigate to Documents and create a new folder:

cd ~/Documents
mkdir myresearch
cd myresearch
git init

git init turns the folder into a git repository. This gives you version control — you can undo changes and track what Claude did. It takes one second and costs nothing.

B2. Create start.md

start.md is your project's first document. It tells you — and Claude — what this project is about.

Open PowerShell and type:

claude

Now ask Claude to help you create start.md:

I want to create a start.md for my research project.
Ask me 5 questions about my research, then write the file.

Answer Claude's questions. It will create start.md in your folder.

Alternatively, create it yourself using this template — open any text editor (Notepad, Obsidian, VS Code) and save it as start.md in your myresearch folder:

# My Research Project

## What this project is
[2–3 sentences: your research topic and main question]

## Current focus
[What you are working on right now — a specific paper, chapter, or dataset]

## What I want Claude to help with

- [task 1 — e.g. reading and synthesising sources]

- [task 2 — e.g. giving feedback on drafts]

- [task 3 — e.g. helping structure arguments]

## What I do NOT want Claude to shape
[1–2 things that should remain your own judgment — e.g. interpretive conclusions,
the research question itself, evaluation of primary sources]

B3. Run the improvement dialogue

If you haven't already opened Claude Code in your project folder:

claude

Then:

Read start.md.
Ask me 5 questions that would help you be more useful to this project.

Answer Claude's questions honestly. After you answer, ask:

Based on my answers, is there anything I should add or change in start.md?

Claude is now oriented to your project. This will make every subsequent session noticeably more relevant.

B4. Add PDFs to your project

Copy 2–3 PDFs into your myresearch folder. You can use:

The PDFs just need to be in the folder — you don't need to open them.

B5. Convert PDFs to markdown

PDF files are hard to work with directly. Markdown files are much better: Claude can read them faster, search them, and cite specific lines.

Ask Claude to convert your PDFs:

Convert all PDF files in this folder to markdown.
Save each one as [original-filename].md

Wait while Claude works through the files. You will see them being created — you can check with dir or in your file browser.

Note: For complex PDFs with tables, equations, or heavy formatting, the conversion will be imperfect. The markdown version is a working document, not a perfect transcript.

B6. Ask substantive questions about your documents

Now ask Claude about the content. Use these as starting points and adapt them to your actual documents:

Overview:

Read all the markdown files in this folder.
What are the main arguments or findings across these documents?

Connections:

What themes or questions appear in more than one of these documents?
Where do the documents agree, and where do they disagree or complement each other?

Gaps and next steps:

Based on these documents, what seems to be missing?
What would I need to read next to build on this?

Project-specific (adjust to your context):

I am working on [your research question].
How do these documents relate to that question?
What do they add, and what do they leave open?

B7. Create a reading note

Ask Claude to produce a structured note from the session:

Create a file called reading-notes.md with:

1. The main argument or finding from each document (1 sentence each)
2. Themes that appear across more than one document
3. Three open questions this reading raises
4. Two things I should investigate next

Use the actual document filenames as headings.

After Claude creates the file: open it in Obsidian or a text editor and read it. Edit anything that is wrong, incomplete, or doesn't reflect your actual thinking. This editing step is part of the exercise — Claude's note is a scaffold, not a finished product. The parts you rewrite are the parts you own.

B8. Optional: Create CLAUDE.md

If you have 5 minutes and want to take the project to Level 2:

I want to create a CLAUDE.md for this project.
Ask me 5 questions about my research, then draft the file.
Save it as CLAUDE.md

Read the draft. Check that it accurately describes your project. Fix anything wrong.

Once CLAUDE.md exists, every future Claude Code session in this folder will open with full project context — Claude will know who you are and what you are working on without re-explanation.

This is the Level 2 transition. A 30-minute investment that pays back on every subsequent session.

→ For what to do next with CLAUDE.md: C.leveling-packages (the 1→2 package)

What you now have

At the end of this tutorial:

This is a working Level 2 or 3 setup. Come back to it, continue the conversation, and add more documents as your project develops.

What to do next

Today: Read C.leveling-packages — specifically the 1→2 package. The 30-minute project description exercise is the single most useful thing you can do after today's workshop.

To go deeper: Claude Code in Action — Anthropic Academy is a free video course from Anthropic with hands-on walkthroughs of real Claude Code workflows. Good follow-up once you have today's setup working.

This week: Bring Claude Code into one real task — reading a paper, giving feedback on a draft, processing a set of documents. Do it with your actual project, not a test case.

When you are ready: Read C.tasks.md to see what becomes possible at Level 3, and C.why.md for when you need a reason to invest more time.

Related