Cookie Consent

We use cookies to enhance your browsing experience, serve personalised ads or content, and analyse our traffic. Learn more

AI in ASIA
Back to Guides
learn
intermediate
Claude

Claude Projects: How to Organise Your Work with AI

Set up Claude Projects to create persistent AI workspaces that remember your files, follow your rules, and stay on-brief across every chat.

9 min read30 March 2026
claude
claude-projects
ai-productivity
anthropic
ai-workspace
productivity
learn
tutorial
Dark cinematic workspace with organised notebook, index cards, and pen on a charcoal surface with blue accent lighting

Claude Projects are persistent workspaces on Claude.ai where your documents, custom instructions, and conversation history are saved and applied automatically to every new chat inside that project.

Each project supports a knowledge base of uploaded files, and Pro, Max, Team, and Enterprise plans extend the effective context using Retrieval Augmented Generation, scaling well beyond the 200,000-token base limit.

As of early 2026, basic Projects access is available on the free Claude.ai plan, making this a feature that professionals across Asia can use regardless of subscription tier.

Why This Matters

Starting each AI conversation from scratch is one of the biggest time drains in using AI for real work. You paste in your brand guidelines, re-explain the project, set the tone again - and still get responses that drift from what you actually want. This is the default Claude experience for most people, and it is fixable.

Claude Projects are workspaces that hold context permanently. Custom instructions tell Claude how to behave: what tone to use, what to avoid, what goals to serve. A knowledge base holds your uploaded documents so Claude can reference them without being told. Chat history persists, so every new conversation in a project builds on previous ones rather than resetting.

For professionals managing multiple clients, content streams, or research workstreams across markets like Singapore, Indonesia, and India - where AI adoption is accelerating - Projects are the difference between Claude as a novelty and Claude as a reliable work tool. Anthropic made basic Projects available to free-tier users in early 2026, significantly expanding who can benefit from this feature across Asia's growing AI user base.

How to Do It

1

Understand what a Project actually is

A Claude Project is a dedicated workspace on Claude.ai with three components working together. First, project instructions: a custom system prompt that applies to every chat in the project, defining tone, rules, goals, and context. Second, a knowledge base: a set of uploaded files (PDFs, Word documents, text files, code) that Claude can reference throughout all conversations. Third, persistent chat history: conversations saved within the project so context carries forward across sessions. Think of it as a configured version of Claude built specifically for one purpose: a client account, a research topic, a writing voice, or a business function.
2

Create your first Project

Log into Claude.ai and look for the Projects section in the left-hand sidebar. Click New Project and give it a specific, descriptive name such as Client: TechCorp Singapore or Content: Newsletter rather than something vague like Work Stuff. A good naming convention from the start saves confusion when you have five or six active projects running in parallel. Treat each project like a folder with a defined purpose, not a chat thread.
3

Write effective project instructions

This is the most important step and the one most people rush. Project instructions are the custom system prompt for your workspace - Claude reads them at the start of every conversation. Cover four things: who Claude is acting as in this project, the tone and style expected, what to avoid, and the primary tasks this workspace handles. Be specific. Vague instructions produce vague outputs. Spend 15-20 minutes writing these properly the first time - it pays back quickly.
4

Build your knowledge base

Inside your project, upload the documents Claude should reference. Common examples: brand guidelines, past campaign reports, product briefs, client contracts, competitor analysis documents, or research you want Claude to draw from consistently. Free plan users can upload files but with limited RAG scaling. Pro, Max, Team, and Enterprise plans use Retrieval Augmented Generation, meaning Claude intelligently retrieves relevant parts of your documents rather than loading everything at once. Keep your files current: if a brand guideline changes, replace the old file immediately.
5

Use your Project across multiple chats

Once configured, every new chat you start inside the project automatically loads your instructions and knowledge base. You do not need to paste anything in - just describe the task. The real benefit emerges across sessions: if you ask Claude to draft a campaign brief today, then come back three days later to refine the tone, Claude has the history of that prior conversation available. Start new chats for distinct tasks rather than continuing one very long thread. This keeps history manageable and ensures Claude is not retrieving irrelevant earlier context.
6

Manage and maintain your Projects over time

A project is only as useful as its maintenance. Review your instructions every four to six weeks and refine them based on what has been working and what has been producing off-target responses. Add new files to the knowledge base as the project evolves. Archive or delete projects you no longer use so your sidebar stays clean. For Team plan users, projects can be shared with collaborators - meaning your whole team works from the same instructions and knowledge base. This is particularly powerful for agencies and small businesses managing multiple markets or languages.

What This Actually Looks Like

The Prompt

We need to refresh the monthly newsletter intro for March. Last month felt a bit stiff. Use the brand voice from the guidelines and reference the Singapore AI adoption stats we uploaded.

Example output — your results will vary based on your inputs

A newsletter introduction written in the established brand voice, drawing directly on the uploaded statistics file, matching the tone set in the project instructions without the user needing to paste in any background.

How to Edit This

The output will directly reflect whatever tone you have defined in your instructions. If the result is still too formal, refine the project instruction rather than reprompting - the fix belongs in the system prompt, not the individual chat.

Prompts to Try

Set up your project instructions from scratch

I am creating a Claude Project for [describe your use case]. Based on what I tell you, help me write project instructions that cover: the role Claude should play, the tone and style to use, what to avoid, and the primary tasks this project will handle. Ask me the questions you need.

What to expect: Claude will ask targeted clarifying questions and produce a draft instruction set you can paste directly into your project settings.

Brief Claude on a new uploaded document

I have just added [document name] to this project knowledge base. Please read it and give me: a 5-sentence summary of the main points, any facts or figures I should reference regularly, and anything that contradicts or updates what you already know about this project.

What to expect: A structured briefing on your new file, helping Claude integrate it into the project context effectively from the first chat.

Confirm consistency before starting a deliverable

We are about to produce [a campaign brief / three blog posts / a set of email templates] for this project. Before we start, confirm: what is the tone we are using, who is the target audience, and what are the three things every piece of output must reflect based on the project guidelines?

What to expect: Claude recites the key parameters from your project instructions, giving you confidence that outputs will stay on-brief before you invest time in the work.

Identify gaps in your knowledge base

Based on the files in this project and the work we have been doing together, what information is missing that would make your responses more accurate or useful? Be specific about what documents or data you would find helpful.

What to expect: A prioritised list of knowledge base gaps you can then fill, improving response quality progressively over time.

Improve your project instructions based on usage

Look at our last 10 conversations in this project. Based on where my outputs have been off-brief or where I have had to correct you, suggest specific changes to the project instructions that would prevent those issues going forward.

What to expect: Targeted suggestions for refining your instructions based on real usage patterns rather than guesswork.

Common Mistakes

Writing vague project instructions

Instructions like be helpful and professional give Claude almost nothing to work with. The more specific your instructions - tone, audience, format preferences, what to avoid - the more consistently useful your outputs will be. Treat this like writing a job brief for a contractor.

Uploading outdated files and not replacing them

If your knowledge base contains a brand guideline from six months ago, Claude will work from that version. Any response referencing those files will be based on stale information. Set a recurring reminder to audit your knowledge base every few weeks.

Using one Project for everything

Mixing unrelated work streams in a single project pollutes the context and makes instructions impossible to write well. Create a separate project for each distinct purpose: one per client, one per content vertical, one per research topic.

Ignoring the persistent chat history

One of the core benefits of Projects is that Claude retains conversation history across sessions. If you keep opening new chats for every single task without referencing prior work, you are missing the point. Explicitly reference earlier conversations when they are relevant.

Expecting Projects to fix poor prompts

Project instructions set the context and rules, but they do not replace the need for clear, specific prompts within each chat. Projects make Claude more consistent; good prompting makes Claude more precise. You need both.

Tools That Work for This

Claude.ai (Free)

Basic Projects access with file uploads and a 200,000-token context window. Good for getting started and testing the feature without a paid commitment.

Limited RAG scaling on the knowledge base; advanced context extension requires a paid plan.

Claude.ai Pro / Max

Full Projects functionality including RAG-powered knowledge base scaling. The right tier for solo professionals doing sustained, multi-session AI-assisted work.

No collaboration features; projects are personal only.

Claude.ai Team

Adds the ability to share projects with collaborators, making it the right choice for agencies, small teams, and businesses running shared workflows.

Per-seat pricing makes it more expensive to scale to larger teams.

Google Docs

Easy way to prepare and version your knowledge base documents before uploading. Real-time collaboration means multiple contributors can update source files without version confusion.

Must be downloaded as PDF or DOCX before uploading to Claude Projects.

Notion

Useful for organising and storing the outputs Claude produces across project chats - particularly briefs, reports, or content drafts that need to be shared beyond Claude.

No direct integration with Claude Projects; outputs require manual copy-paste.

Smallpdf or Adobe Acrobat

Helpful for combining multiple related documents into a single upload, reducing file clutter in your knowledge base.

Free PDF tools often have page or file size limits that can be frustrating with large documents.

Frequently Asked Questions

Yes. As of early 2026, basic Projects access is available to all Claude.ai users including the free tier. Free plan users can create projects, write instructions, and upload files. The main difference is that the RAG-powered knowledge base scaling is limited compared to paid plans.
Anthropic has not published a hard cap on the number of projects you can create. In practice, keeping five to ten active projects is manageable before the sidebar becomes hard to navigate. Archive or delete projects you are no longer actively using.
Project sharing is available on Claude.ai Team and Enterprise plans. Free, Pro, and Max plan users cannot share projects with collaborators - projects remain personal workspaces on those tiers.
Claude supports PDFs, Word documents, plain text files, code files, and similar text-based formats. Image files can be uploaded within individual chats but are not stored persistently in the project knowledge base.
Claude retains chat history within a project, but very long histories may be partially truncated as the context window fills. For critical reference information, always store it in an uploaded document in the knowledge base rather than relying on chat history alone.

Next Steps

Now you know how to configure a Claude Project, the most effective next step is to build one around your most repetitive AI workflow - the task you find yourself re-explaining to Claude most often. For a broader introduction to what Claude can do beyond Projects, see our guide on How to Use Claude at https://aiinasia.com/guides/learn/how-to-use-claude-anthropic-guide. For context on why AI proficiency is becoming a professional essential across the region, our piece on Southeast Asia enterprise AI adoption at https://aiinasia.com/news/southeast-asia-enterprise-ai-adoption-mckinsey-edb-2026 is worth reading alongside this.

No comments yet. Be the first to share your thoughts!

Leave a Comment

Your email will not be published