Every freelancer knows the ritual: open a new AI chat, paste the same background for the tenth time — who the client is, what the brand voice sounds like, what you decided last Tuesday. It’s dead time, and it leaks quality, because the AI never quite has the full picture.
Claude Projects kills that ritual. A Project is a dedicated workspace inside Claude built around three things: custom instructions (rules Claude follows in every chat), Project Knowledge (your uploaded files — briefs, style guides, past deliverables), and persistent chat history. Every new conversation inside the Project starts with Claude already knowing your context. No more re-explaining.
Setting up your first client Project
Step 1: Create the Project. In Claude’s sidebar, click Projects, then “Create Project.” Name it after the client or engagement — “Acme Dental — Content Retainer” beats “Project 3.”
Step 2: Write real custom instructions. This is the highest-leverage field in the whole feature, and most people waste it on “be helpful and professional.” That’s a wish, not an instruction. Instead, write the brief you’d give a new contractor:
- Who the client is and who their audience is
- The deliverable formats you produce (with examples)
- Tone rules with do/don’t pairs: “Write like a confident peer, never like a textbook. Never use ‘delve,’ ‘tapestry,’ or ‘game-changer.’”
- Length and structure defaults: “Blog drafts: 1,200 words, intro under 80 words, H2s as questions”
- What to ask before acting: “If a request is ambiguous, ask one clarifying question instead of guessing”
Anthropic’s own docs note that the project name and description are not treated as behavior instructions — the real behavior lives in the “Set project instructions” field and your knowledge files. Put the substance there.
Step 3: Upload the knowledge base. Click “Add content” in the Project Knowledge area. For a client Project, upload: the brand style guide, 2-3 examples of approved past work, the client brief or contract scope, audience personas, and any reference docs (pricing pages, product one-pagers). Claude accepts PDFs, Word docs, markdown, text, images, and code files. Favor fewer, well-curated files — a bloated knowledge base eats into the context budget every turn.
Step 4: Star it and start working. Star the Project for quick access. Open a new chat inside it and test: “Draft a blog outline for next week using the content pillars in the knowledge base.” If Claude reaches for your files without being asked, the setup is working.
The freelancer Project stack
Don’t stop at one. The pros run a small portfolio of Projects, one per context:
- Per-client Projects. Each retainer client gets one. Knowledge: their brand docs, past deliverables, meeting notes. Instructions: their voice, their formats, their approval quirks.
- A “Business OS” Project. Your own operations manual: service descriptions, pricing, proposal templates, your bio, testimonial bank. Use it to draft proposals, scope documents, and follow-up emails in your voice.
- A “Content Engine” Project. Your publishing system: niche definitions, headline formulas that worked, SEO checklists, repurposing workflows. Feed it your analytics summaries monthly so its advice stays grounded in what actually performed.
- A “Learning” Project. Course notes, book highlights, frameworks you’re studying. Ask it to quiz you or apply a framework to a live client problem.
Projects are available on all Claude plans — free accounts can create up to 5, which is honestly enough to run the stack above. Paid plans raise the limits.
The 200K context window is your unfair advantage
Each Project carries a large context window (200K tokens at launch — roughly a 500-page book), which means you can load serious reference material: an entire client website export, a quarter of meeting transcripts, a full brand book. Use it. The freelancers getting the most from Projects treat the knowledge base as a living client brain: after every engagement, drop in the final deliverables and a short “what worked” note. Six months later, that Project knows the client better than you remember.
Three mistakes that ruin Projects
- Vague instructions. “Be professional” does nothing. Specific rules with examples do everything.
- Dumping everything in. Uploading 40 random files makes every response slower and dumber. Curate: if a file hasn’t been useful in a month, remove it.
- One mega-Project for everything. Contexts bleed. A client Project that also contains your personal journal produces weird, off-voice output. Separate Projects for separate contexts.
What changed in 2026
Anthropic redesigned Projects in September 2026 around ongoing work: parallel conversation threads, shared memory across threads, and longer project instructions. The direction is clear — Projects are becoming persistent workspaces, not just folders. Build the habit now and you’ll be fluent by the time the next capabilities land.
Official sources: What are Projects? — Claude Help Center · Collaborate with Claude on Projects — Anthropic
The bottom line
Claude Projects turns scattered AI chats into organized client infrastructure. One Project per client, one for your business, one for content — each with sharp instructions and a curated knowledge base — and you stop paying the “re-explaining tax” forever. Setup takes an hour per Project. The payoff is every hour after that.