Claude Projects: Set Up a Persistent Workspace

Fast Forward · Episode 01

Claude Projects: Build a Persistent AI Workspace

If you use Claude regularly for work, you should not have to explain your role, project, tone and priorities every time you start a new chat. This Fast Forward episode shows how Claude Projects can turn repeated setup into a reusable working context.

What this Fast Forward episode is about

If you use Claude regularly for work, there is a good chance you are wasting the first few minutes of every session explaining the same things: your role, your project, your preferred tone, the current context and what matters right now.

Claude Projects helps solve that by giving you a dedicated workspace for a specific area of work. Instead of starting cold every time, conversations inside the Project can begin with the right context already in place.

The point is simple: one project, one briefing, fewer repeated explanations.

Why this matters

AI feels much more useful when it understands the frame of the work. Without that frame, you spend time correcting tone, restating goals and explaining constraints that should already be known.

A persistent workspace is especially useful for client work, writing projects, analysis, reporting, planning and any activity where the same context matters across multiple conversations.

Problem Every chat starts cold

You keep repeating the same background before Claude can do useful work.

Shift Context becomes reusable

Project instructions and uploaded documents create a stable starting point for future conversations.

Impact Better first drafts

Claude can start closer to the right tone, structure and level of detail.

Mindset AI as workspace, not chat box

The value increases when the tool is organised around real work contexts.

What Claude Projects actually is

A Project is a dedicated workspace inside Claude.ai. It gives Claude a clearer working environment for a specific subject, client, engagement, writing project or area of responsibility.

Inside the Project, you can set instructions, upload documents and keep the related conversation history together. The result is a more persistent working context.

Part 1 Project Instructions

A written briefing that tells Claude who you are, what the project is and how you want it to work with you.

Part 2 Uploaded documents

Style guides, briefs, strategy notes, reports or reference material Claude can use within that Project.

Part 3 Conversation history

Related chats stay in the Project so your work does not scatter across unrelated conversations.

Result Less setup time

Each new conversation can start closer to the right context, rather than beginning from scratch.

The five-step setup

The setup is simple. The quality comes from how clearly you describe the context and how well you maintain it over time.

Step 1 Open Claude.ai

Find Projects in the left sidebar and create a new Project.

Step 2 Name it clearly

Use a client name, engagement name or subject area, such as Quarterly Reporting or My Writing Projects.

Step 3 Write Project Instructions

Brief Claude like a capable contractor starting work with you today.

Step 4 Upload relevant documents

Add the reference material Claude should always have available inside that Project.

Step 5 Start inside the Project

Use that workspace for the relevant work so your chats stay grouped and contextually useful.

Rule One Project per context

Do not mix unrelated work. Separate contexts keep Claude’s guidance cleaner and more useful.

Project Instructions template

The Project Instructions are the most important part. Keep them short, specific and practical. This starter template gives Claude the key context without overwhelming it.

I am a [role] working on [project or engagement name].My main goal with this Project is [outcome].I want you to [tone guidance].Key things to know: – [Specific context point 1] – [Specific context point 2] – [Specific context point 3]When drafting, default to [format preference].Flag uncertainty rather than guessing.

Best use: write instructions as if you are briefing a capable contractor who needs to be productive quickly.

What makes good Project Instructions

Good instructions do not need to be long. They need to be specific enough to improve the first response and clear enough to guide future work.

Principle 1 Specific beats vague

“Senior business analyst working on a membership platform for an Australian education organisation” is stronger than “analyst”.

Principle 2 Include output preferences

Tell Claude whether you want plain prose, bullets, tables, pushback, concise drafts or detailed analysis.

Principle 3 Add constraints

Include tone rules, audience expectations, source handling, formatting needs or things to avoid.

Principle 4 Keep it updated

As the work changes, update the instructions so Claude stays aligned with the current project reality.

When to use Claude Projects

Claude Projects are most useful when you return to the same context repeatedly. They are less important for one-off questions where the setup would take longer than the task.

Good fit Client or project work

Use a Project when the same client, engagement, constraints and documents come up repeatedly.

Good fit Writing and content

Store tone rules, content strategy, examples and audience guidance in one place.

Good fit Analysis and reporting

Keep reports, source documents, templates and analytical rules together.

Poor fit One-off tasks

For simple questions or throwaway tasks, a normal chat may be enough.

The practical payoff

Once a Project is set up well, every conversation inside it starts at a higher level. You get better first drafts, fewer corrections and less time spent repeating context Claude should already know.

The first setup may take a few minutes, but the payoff compounds every time you return to the same workspace. That is the real value: not a fancy feature, but less repeated setup and more useful work.

Claude Projects turn repeated context into reusable leverage.

What this prepares you for

This Fast Forward episode introduces a wider Zero to AI pattern: AI works better when you create structured workspaces, not scattered conversations.

That pattern connects directly into assistants, personas, panels and the later work around AI-supported operating rhythms.

Stop repeating the setup.

Build one clear Project, give Claude the right briefing and let your workspace carry the context forward.

' + ''; }fields.forEach(function (field) { field.addEventListener('input', function () { saveField(field); renderAsset(); updateProgress(); showToast('Saved'); });field.addEventListener('change', function () { saveField(field); renderAsset(); updateProgress(); }); });pathwayInputs.forEach(function (input) { input.addEventListener('change', function () { setPathway(input.value, true); }); });pathwayButtons.forEach(function (button) { button.addEventListener('click', function () { setPathway(button.getAttribute('data-pathway-choice'), true); }); });const copyPromptButton = lab.querySelector('#z2ai-copy-prompt'); const promptText = lab.querySelector('#z2ai-prompt-text');if (copyPromptButton && promptText) { copyPromptButton.addEventListener('click', function () { copyText(promptText.textContent.trim(), 'Prompt copied', 'Copy failed', showToast); }); }const copyAssetButton = lab.querySelector('#z2ai-copy-asset');if (copyAssetButton) { copyAssetButton.addEventListener('click', function () { copyText(getAssetText(), 'Asset copied', 'Copy failed', showToast); }); }const printAssetButton = lab.querySelector('#z2ai-print-asset');if (printAssetButton) { printAssetButton.addEventListener('click', function (event) { event.preventDefault();renderAsset();const printWindow = window.open('', '_blank', 'width=900,height=1100');if (!printWindow) { showToast('Print window blocked'); return; }printWindow.document.open(); printWindow.document.write(buildPrintHtml()); printWindow.document.close(); printWindow.focus();window.setTimeout(function () { printWindow.print(); }, 350); }); }const clearLabButton = lab.querySelector('#z2ai-clear-lab');if (clearLabButton) { clearLabButton.addEventListener('click', function () { const confirmed = window.confirm('Clear the saved responses for this Learning Lab in this browser?'); if (!confirmed) return;window.localStorage.removeItem(storageKey);fields.forEach(function (field) { field.value = ''; });setPathway('minimum', false); renderAsset(); updateProgress(); showToast('Responses cleared'); }); }restoreState(); }labs.forEach(initLab); })();