Intermediate

Project Use Cases

Projects work best for ongoing, context-heavy tasks that span multiple sessions. Here are five well-tested patterns for getting the most out of Claude Projects.

Code Project Assistant

Setup

  • Knowledge base: Upload your codebase (or the most relevant parts), architecture documentation, README, and any standards documents (linting rules, naming conventions)
  • Custom instructions: Specify the language, framework, coding conventions, and what Claude should assume about the stack. E.g. "This is a Next.js 14 App Router project using TypeScript and Tailwind CSS. Use server components by default unless client-side interaction requires otherwise."

What it unlocks

Claude answers questions about the codebase with actual knowledge of your specific code — not generic examples. It writes new functions consistent with your patterns. It explains decisions in context. Much more useful than asking generic coding questions in a plain conversation.

Research Project

Setup

  • Knowledge base: Upload the key papers, reports, and sources you're working with. Add a running notes document that you update with key insights.
  • Custom instructions: Define the research domain, your current focus question, and how you want Claude to engage ("Challenge my assumptions, not just confirm them. Flag when claims lack support.")

What it unlocks

Ask synthesis questions across multiple papers: "What do these sources collectively say about [topic]?" Build on prior conversations — "Following on from our discussion of the methodology gap yesterday...". Claude tracks the developing picture of your research without you re-loading context each session.

Client Work

Setup

  • Knowledge base: Client brief, brand guidelines, prior deliverables, meeting notes, feedback received
  • Custom instructions: Client name, their industry, their audience, tone preferences, and any "do not mention" constraints (competitors, sensitive topics)

What it unlocks

One Project per client. Every piece of work for that client is context-aware — consistent voice, consistent assumptions, no repetitive re-briefing. When a client gives feedback, add it to the knowledge base and all future work reflects it.

Writing Project

Setup

  • Knowledge base: Your style guide (or past writing samples), an outline, prior draft versions, notes on what you want to change
  • Custom instructions: Your voice (or a description of it), any rules ("never use passive voice", "Oxford comma always"), the target audience, and where you are in the process

What it unlocks

Claude maintains your voice across drafts. You can pick up mid-draft without re-briefing. Claude remembers that you rejected a particular approach in session 3 and won't re-suggest it. Feedback from a reviewer can be added to the knowledge base so future drafts automatically incorporate it.

Personal AI Assistant

Setup

  • Knowledge base: A personal context document you maintain: your role, projects you're working on, preferences Claude should know, standing goals
  • Custom instructions: Your communication preferences, what kind of help you typically want, and any defaults ("always ask for the desired output format if unclear before proceeding")

What it unlocks

A truly personalised assistant that knows your context, not a fresh start every time. Update the context document when your situation changes (new job, new project, new goals). Over time, this Project becomes the highest-value single conversation interface you have with Claude.

Tips Across All Use Cases

  • Start with a minimal setup and add knowledge base files as the need becomes clear — don't over-load at the start
  • After your first 3–4 conversations in a new Project, review the custom instructions and refine based on what Claude is getting right and wrong
  • For long-running Projects, periodically add a "context update" document to the knowledge base summarising key decisions and current state
  • Name Projects clearly so you can find them quickly in the sidebar — "Client: Acme Corp" is more useful than "Work Stuff"

Checklist: Do You Understand This?

  • Code project: upload codebase + conventions, specify language/framework in instructions — enables context-aware code help
  • Research project: upload papers and notes, define your research question — enables cross-source synthesis with continuity
  • Client work: one Project per client with their brief and brand guidelines — enables consistent, context-aware deliverables
  • Writing project: upload drafts and style guide, define your voice — enables continued drafting without re-briefing
  • Personal assistant: maintain a personal context document, specify your preferences — builds a persistent, improving assistant over time

Page built: 01 Jun 2026