Lab 4 Tutorial • Estimated time: 30 minutes

Stop Starting Every Conversation from Scratch

If you’ve used Claude for more than a few days, you’ve probably noticed something frustrating: every time you open a new chat, Claude has no idea who you are or what you’re working on. You explain your project, share your context, maybe upload a document — and then the next day you do it all over again. That’s not how a good working relationship should feel. 

Projects and Artifacts are Claude’s answer to this. Projects give you a persistent workspace — with standing instructions and a knowledge base — that every conversation inside it automatically inherits. Artifacts are standalone outputs that live in your sidebar, beyond any single chat, ready to be refined and reused. 

This walkthrough of Lab 4 in our Working Smarter with Claude course will show you how both works, using a realistic scenario so you can see it in action, not just in theory.

---

What You’ll Walk Away With

By the end of this tutorial, you’ll know how to: 

• Create a Project and write standing instructions that shape how Claude behaves inside it
• Upload reference documents so Claude works from your material, not general knowledge 
• Verify that context carries automatically across multiple conversations in the same Project 
• Share a Project with teammates (Team and Enterprise plans) 
• Build an Artifact — a reusable, interactive output that lives in your sidebar, not just in a chat
  

---

The Scenario: A Product Launch Program Manager

You have a General Availability (GA) launch date on the calendar, a cross-functional team to coordinate, a Go/No-Go decision coming up, and a lot of documents to manage. 

You’ll be doing Claude conversations for the next 90 days. Without Projects, every conversation starts cold. You would have to re-explain what the product is, what role you’re playing, what tone you need every single time. 

With a Project? You set it up once. Every conversation inside it inherits your context automatically. Let’s build that.

---

Step 1: Create Your Project

Projects and click + New Project. 

Give it a descriptive name tied to your actual work. Something like “Pulse Launch — May 5 GA” tells you exactly what this workspace is for. Add a short project description if you’d like — this is just for your own organization. Claude doesn’t read the project name or description. It only reads the Project Instructions you’re about to write.

Why this matters: Without a recognizable name, you’ll end up with “Project 1,” “Project 2,” and “Untitled.” Name it like a folder on your desktop that you actually need to find later.

---

Step 2: Write Your Project Instructions

Instructions area and click to edit it. What you write here is injected at the start of every conversation in this project — it’s Claude’s standing brief. Think of it as the memo you’d hand to a new assistant on their first day. Here’s what good project instructions look like:  

Notice what this does: it establishes Claude’s role, sets the tone, specifies the output format, and tells Claude where to look for facts.

---

Step 3: Upload Your Reference Documents

Files section. This is your Project Knowledge Base. 

Upload the documents Claude should draw from in every conversation. For a product launch, that might be: 

• The overall launch plan 
• A RACI matrix (who’s responsible for what decisions and actions) 
• A Go/No-Go readiness checklist 

 Once uploaded, those documents are available to Claude in every conversation inside this project. No re-uploading, no pasting content. Claude just has them.

Files section before moving on. You should see thumbnails and filenames listed there.

A Note on Project Memory 

Memory section in the right-side panel — right above the Files section — that says “Project memory will show here after a few chats.” This isn’t a problem. It just means Claude hasn’t had enough conversations inside the project yet to start building context. Give it a few chats and it will populate on its own. 

It’s also worth knowing that Project memory and your personal memory are two separate things: 

• Personal memory (found in Settings → Capabilities) captures general facts about you — your role, preferences, how you like to work — and carries into every conversation you have across all of Claude 
• Project memory captures patterns and context specific to this project’s conversations — and stays inside the project 
  

Think of personal memory as your résumé. Project memory is more like the running notes a colleague takes as they work alongside you on a specific engagement.

---

Step 4: Start a Conversation and Watch the Context Apply

Inside the project — not a new standalone chat. Look for a New Chat button within the project view. Try a prompt that can only be answered well if Claude has absorbed your documents:

If your project is set up correctly, Claude will: 

• Answer from the actual documents you uploaded, not generic knowledge 
• Use the tone and format you specified in your instructions 
• Name the accountable owners from your RACI, not invent placeholder names 

 If Claude gives a vague or generic answer that doesn’t reference your documents, check two things: (1) are the files uploaded to the project? and (2) do your instructions tell Claude to use them?

---

Step 5: Test Context Persistence Across Conversations

Let’s have a second conversation inside the same project — click New Chat again, still within the project. Now ask something completely different, without re-explaining your project:

Claude should answer as if it already knows the full context of your work — because it does. The project instructions and uploaded files are still there. You didn’t have to re-explain anything. 

This is the moment when Projects stop feeling like a feature and start feeling like a working relationship.

---

Step 6: Share Your Project with Teammates

If your organization is on a Team or Enterprise plan, Projects aren’t just personal workspaces — they’re collaborative ones. You can share a project with specific colleagues or make it available to your entire organization. 

Share project button next to the project name. Add teammates by name or email, then choose their permission level: 

• Can use — they can read the instructions, access the knowledge base, and start conversations inside it, but can’t change anything 
• Can edit — they can modify the instructions, update the knowledge base, and manage member access 

 Public when creating it. It becomes discoverable to everyone in your org under the Team tab in the Projects section.

Step 7: Build an Artifact

A chat response disappears into your history. An Artifact lives on. 

Artifacts are standalone outputs — interactive tools, dashboards, documents, calculators, checklists — that Claude generates and stores in your sidebar. You can return to them anytime, refine them across multiple sessions, and share them outside of Claude entirely. 

For a launch program manager, a natural candidate is an interactive Go/No-Go readiness scorecard — something you can use to track whether you’re ready to launch. 

How to create it: 

1. In the left sidebar, click Artifacts under the main navigation. 
2. Click New Artifact — a category picker will appear. Choose something like a tracker, checklist, or dashboard. If nothing fits, click Build from scratch. 
3. The Artifact creation flow opens its own conversation. Because this is separate from your project, your uploaded files aren’t automatically available here — upload your readiness criteria document again using the + button. 
4. Submit a prompt like the one below. 
   

Artifacts with a name. That’s your persistent, reusable tool — not a chat response, but an actual thing you made.

---

Step 8: Refine and Explore Version History

Artifacts aren’t one-and-done. You can keep improving them, and Claude tracks every version. In the Artifact’s conversation panel, try requesting a refinement:

Version selector. You can navigate back to the original and forward to the new one, compare them, and revert if you prefer the earlier version. This is version control for the outputs you build with Claude.

---

The Mental Model That Makes This Click

Once you’ve done all of this, here’s how to think about the three layers:

Layer	What it is	What it’s for
Project Instructions	Your standing brief for Claude	Tone, role, format, and rules — set once, applied always
Knowledge Base	Your uploaded documents	The facts Claude draws from in every conversation
Artifacts	Standalone outputs	Reusable tools and documents that live beyond any single chat

Standalone chats are great for quick questions. Projects are for ongoing work. Artifacts are for outputs you’ll use again.

---

Common Beginner Mistakes (and How to Avoid Them)

Uploading documents but writing vague instructions. 

The documents give Claude raw material. The instructions tell it what to do with that material. You need both. 

Creating Projects for one-time tasks. 

Projects earn their value through repeated use. If you’ll only have one conversation about something, a regular chat is fine. If you’ll have ten conversations, build a Project.

Treating Artifacts like chat responses. 

Chat responses scroll away. Artifacts are persistent, versionable, shareable outputs. If Claude builds something useful, turn it into an Artifact. Don’t let it get buried in a chat thread. 

Forgetting to start conversations inside the project. 

If you click New Chat from the main sidebar instead of from within the project, you get a standalone chat with no project context. Always start new conversations from within the project itself. 

Worrying when Project memory looks empty. 

This is normal. Project memory populates gradually over several chats — it’s not something you configure, it builds on its own. Personal memory and Project memory are separate; don’t confuse the two.

---

Knowledge Check

Try to answer each question before reading the answer.

Question	Answer
What’s the difference between Project Instructions and the Project Knowledge Base?	Instructions tell Claude how to behave — tone, format, role. The Knowledge Base gives Claude what to know — your actual documents and source material.
You’re starting your third conversation about the same product launch. Where should you start it?	Inside the project — so it automatically inherits the instructions and knowledge base without re-explaining anything.
You built a useful status tracker in a chat response. What should you do with it?	Rebuild it as an Artifact so it lives in your sidebar, can be refined over time, and shared beyond the chat.
You want to iterate on an Artifact without losing your original version. What happens?	Claude creates a new version and keeps the old one. Use the version selector at the top of the Artifact panel to navigate between them.
You’re on a Pro plan and want to share your project with a colleague. Can you?	Not with a Pro plan — sharing is only available on Team and Enterprise plans. On Pro, Projects are personal.

---

One Last Thought

Most people use Claude like a search engine — type a question, get an answer, move on. That works fine for simple tasks. But serious work is ongoing. It has context, history, stakeholders, and documents.

 Projects and Artifacts are how Claude handles serious work. Set up a Project for any engagement you’ll return to more than a couple of times. Build Artifacts for outputs you’ll use. You’ll stop feeling like you’re constantly catching Claude up, and start feeling like it’s genuinely working alongside you. 

That’s the difference between a tool you use occasionally and one you rely on. 

Looking for more ways to work smarter with Claude? Our 1-day Working Smarter with Claude Chat course is live, instructor-led, and can be customized to your team’s goals. Contact us to get started upskilling your team.