Claude Projects: your first context workflow (2026)
Claude Projects is the space on claude.ai where instructions and files live with the thread. It is not another chat: it is a container with rules and documents Claude reads on every message. This flow takes 20 minutes and leaves you a reusable project for repeated work — a brief, a client, a document type.
The problem it solves
People paste the same PDF into every new conversation or rewrite the same instructions every Monday. A loose chat does not remember last week’s tone or files. Without a project, you repeat context and Claude invents more freely.
What you will learn
- Decide whether the job deserves a Project or a loose chat
- Create a project and write instructions that survive the thread
- Upload knowledge files Claude consults on every message
- Chat inside the project and verify it uses the context
- Know when to go back to a normal chat
Before you start
- An account on claude.ai (the Free plan is enough to try Projects with limits)
- A real case: a client, a report type or a brief you repeat
- At least one reference file (PDF, doc or exported notes) or a clear instruction paragraph
- 20 minutes
Steps 1
Decide if you need a Project or a loose chat
Before creating anything, answer: will you repeat this work several times with the same rules and documents? For a one-off email, stay in the normal chat. For a weekly brief, a client with a tone guide or a templated report, a Project is worth it. Write the purpose in one line: «Draft monthly reports for Client X with their glossary and direct tone.»
What you should see: A one-line purpose written in a notes app. A clear decision: Project yes or loose chat.
Steps 2
Create the project on claude.ai
Go to https://claude.ai and open Projects in the sidebar (or projects menu). Click «New project». Give it a concrete name — not «Writing» but «Client X Reports» or «Weekly product brief». The name tells you whether you will come back next week.
What you should see: An empty project with fields for project instructions and a file area. The name reflects the use case.
Steps 3
Write the project instructions
In «Project instructions», write what Claude must always remember: role («You are an internal report editor»), tone («Direct, no empty adjectives»), output format («Bullets with owner and date») and bans («Do not invent figures; if data is missing, mark [MISSING]»). Paste a sample paragraph if you have brand voice. These instructions apply to every chat inside the project.
What you should see: An instruction block of 5–15 lines with role, tone, format and at least one explicit ban.
Steps 4
Upload knowledge files
Add the PDFs, docs or notes Claude should consult: client glossary, report template, style guide, past notes. Do not upload the whole drive: 2–5 relevant files. If a file is outdated, remove it. Project files are read on every new conversation inside it.
What you should see: A file list in the project panel with recognizable names. No duplicates or stale docs.
Steps 5
Open the first chat inside the project
From the project, start a new chat — not from Claude’s main screen. Write a concrete task that uses the context: «Using the project template, draft the March report. Use only data in the attached files.» The chat inherits instructions and files without you pasting them again.
What you should see: A thread inside the project with your prompt. Claude mentions or follows the template and tone from the instructions.
Steps 6
Verify it uses the context (and fix if not)
Review the reply: does it follow the format in the instructions? Does it use glossary terms? Does it mark [MISSING] instead of inventing? If something fails, adjust the project instructions — do not repeat the same blind prompt. A «Banned: paragraphs longer than 3 sentences» in instructions beats three correction messages in the chat.
What you should see: A reply aligned with instructions and files, or edited project instructions after spotting a failure.
Steps 7
When to stay in the Project and when to return to a loose chat
Stay in the Project for repeated work on the same case: same client, same doc type, same rules. Return to a loose chat for one-off questions, tone experiments or topics that do not share files. If a second client needs another glossary and tone, create another project — do not mix contexts in one.
What you should see: A clear documented decision: this project = Client X reports; other tasks = normal chat or another project.
Real use cases
Monthly client report
Instructions with tone and format. Glossary and template in files. Each month you open a new chat inside the same project.
Weekly product brief
Structure rules (problem, proposal, metrics). Last week’s notes as reference. One chat per sprint.
Editing with a style guide
Guide PDF in the project. Instructions ban slogans. Each long piece is edited in a new thread without re-uploading the guide.
Common mistakes
Creating a project for a single question
A loose chat is enough. Projects pay off when you repeat context.
Generic instructions
«Be helpful» fixes nothing. Role, format and bans do.
Mixing clients in one project
Two tones and two glossaries in one place confuse. One project per case.
Uploading stale files
Claude reads what you upload. If the template changed, update the file.
Close
Claude Projects wins when you have repeated work with fixed rules and documents. In 20 minutes you leave with written instructions, files in place and a first chat that does not start from zero. If the case is one-off, the loose chat is still the right choice.
Next steps
- To edit a long piece inside the project, use Claude’s section-by-section editing method.
- If you still lack a briefing method, start with ChatGPT first steps and come back here with a repeatable case.
- Create a second project only if you have another client or document type — do not duplicate for sport.
Free guide
Want the 15-tools guide?
We send it free. We will also ping you when the next guide in this path goes out.
Unsubscribe whenever you want. We do not sell the list.