Claude Projects and Artifacts: How to Use Both for Persistent Work

Claude Projects holds your context. Claude Artifacts holds the working output. These are two distinct Claude features that solve different parts of the same problem: making long-running work persist between conversations. Most users start with one and never discover the other. The combination is what unlocks Claude as a workspace, not just a chat. This guide covers what each feature does, what's in the box at Pro versus Max, how to use them together, and where the combination still falls short for power users running cross-tool workflows.

Key Takeaways

• Claude Projects is your knowledge base and instructions container. Files, system prompts, and uploaded documents that persist across every chat inside the project.²
• Claude Artifacts is your live workspace. Interactive apps, documents, and visualizations that render in a dedicated panel, can persist across sessions, call APIs, and connect to MCP servers.¹
• Free users can create up to 5 Projects. Paid plans get unlimited projects, RAG over the knowledge base, persistent Artifacts storage, and MCP integrations.

What Claude Projects actually does

A Claude Project is a container around a single body of work. You name it, give it a system prompt, and upload files to its knowledge base. Every chat you open inside the project starts with those instructions and that knowledge already loaded.²

Setting one up takes under a minute. Navigate to claude.ai/projects and click "+ New Project" in the upper right.³ Give it a name and a description (the description is for you, not Claude). The project knowledge base appears on the right side of the project's main page, with a "+" button for adding files.

What you can upload: PDF, DOCX, CSV, TXT, HTML, ODT, RTF, and EPUB. Individual files can be up to 30MB.³ The total content needs to fit within the context window for free users. Paid plans automatically scale through Retrieval Augmented Generation (RAG) so large knowledge bases stay performant.³

Common patterns that work well:

Book or long writing project. Upload the outline, the existing chapters, the style guide. Set the project instructions to your tone preferences and structural rules. Every new chat inside the project picks up where the last one left off.

Codebase work. Upload the README, the architecture doc, the API spec. Add project instructions covering your conventions. Use it when you don't want to spin up Claude Code but need codebase-aware chat.

Research synthesis. Upload the source PDFs. Set the project instructions to your research questions. Each chat builds on the same source material instead of you re-uploading every session.

Recurring workflow. Job application research, content production, customer Q&A. Anywhere you do the same kind of work multiple times with consistent rules.

Free users can create up to 5 Projects. Paid users (Pro, Max, Team, Enterprise) get unlimited projects plus the RAG layer that handles large knowledge bases.³

What Claude Artifacts actually does

Artifacts is a different thing. Where Projects is the static context layer, Artifacts is the working output layer.¹

When you ask Claude for a document, code, an interactive app, a visualization, or anything substantial enough to want to view separately, Claude creates an artifact: a dedicated panel that opens to the right of your chat. The chat continues on the left. The artifact updates as you iterate. You can ask Claude to modify it, regenerate it, or roll back to a previous version. Edits to prior chat messages create branched versions, each with its own artifact state.

In 2026, Artifacts shipped four major capability upgrades:

Persistent storage (up to 20MB per artifact). Artifacts can save data across sessions. The same artifact you opened last Tuesday remembers its state when you reopen it.¹

Direct API calls. Artifacts can make HTTP calls to external APIs from inside the artifact context. Means you can build small live apps that fetch real data.¹

MCP integrations. Artifacts can connect to Model Context Protocol servers, which lets them read from and write to external tools (Asana, Google Calendar, Slack, Notion, your own MCP server).¹

Live Artifacts that refresh with current data when reopened. A dashboard you built last month pulls the latest numbers next time you open it without you having to re-prompt.

The publishing flow is straightforward: click the publish button at the bottom of the artifact panel, share the generated link. Recipients don't need a Claude account to view, and users with one can remix (create their own editable copy).¹

Using Projects and Artifacts together

The two features compound when used together. The pattern that shows up repeatedly:

The Project holds the source material and project instructions. The Artifact is what you build out of that material in any given chat. Same Project might generate dozens of artifacts over its lifetime (an outline, a draft chapter, a visualization, a calendar, a budget sheet).

Concrete example: a content team running a blog production workflow. The Project knowledge base holds the brand voice doc, the keyword strategy spreadsheet, the past 20 published posts as reference. Project instructions set the tone rules and word count targets. Each new blog post is a chat inside this Project that produces an artifact (the draft). The team iterates inside the artifact panel, ships when done, and starts the next post inside the same Project.

Concrete example for builders: a feature spike Project. Knowledge base holds the codebase README, the relevant architectural doc, the existing API spec. Project instructions cover testing conventions and the team's definition of done. Each spike (one chat inside the Project) produces an artifact (a prototype script, a sketch of the data model, a quick visualization). Persistent storage means you can come back to the prototype next week without rebuilding state.

The thing to watch: artifacts created inside a Project are scoped to that conversation, not the Project as a whole. If you want a useful artifact to be reusable in other conversations within the same Project, drop it back into the knowledge base as a file. Some Claude-generated content (code, drafts, analysis) can be added directly to the Project's knowledge base for future reference.³

What Pro, Max, and Free tiers each get

The plan differences for Projects and Artifacts:

Free. Up to 5 Projects. Artifacts available. No RAG over large knowledge bases. No persistent Artifacts storage. No MCP integrations.

Pro ($20/month). Projects with a higher cap than Free. RAG layer for large knowledge bases.³ Persistent Artifacts storage (up to 20MB per artifact).¹ MCP integrations on Artifacts. Direct API calls from Artifacts. Access to Anthropic's latest Sonnet and Opus models.

Max 5x ($100/month). Everything Pro has, plus 5x the session and weekly usage limits compared to Pro.⁴ Priority access to new features and models, including features like voice mode. Max tiers get the new capabilities first as they ship.

Max 20x ($200/month). Same feature set as Max 5x, with 20x the Pro session and weekly allowance.⁴ Fits teams or heavy developers running multiple parallel agents.

The decision tree most users converge on: if you're using Projects or Artifacts more than occasionally, the jump from Free to Pro pays for itself within the first week. The jump from Pro to Max is about usage limits, not features. Hit Pro's cap a couple of times in a week and Max 5x becomes the right move.

For the full Claude Max upgrade-decision framework, see is Claude Max worth it.

Where Projects and Artifacts fall short

Projects and Artifacts are strong inside Claude. The honest limits show up the moment you step outside.

Project-scoped only. The Project's instructions and knowledge are only visible inside that Project. Start a Project for a book, then switch to a different Project to brainstorm a related side idea, and the second Project doesn't see anything the first one knows.

Inside Claude only. Open ChatGPT to bounce an idea off a different model and ChatGPT doesn't see your Project context. Switch to Claude Code in the terminal and it doesn't read your Claude Projects either. That uses its own CLAUDE.md files instead. See the CLAUDE.md guide for that side of the stack.

Knowledge base is static. The Project knowledge base grows when you upload to it. It doesn't grow automatically from your conversations. If you discover something useful inside a Project chat, you have to manually move it into the knowledge base for the next chat to see it.

Artifacts persistence is per-artifact. A persistent artifact remembers its own state across sessions but doesn't share state with other artifacts in the same Project. Each one is its own bubble.

Cross-tool coverage is zero. This is the big one. Projects + Artifacts together build a workspace inside Claude. Outside Claude, none of it exists.

How to extend with a cross-tool memory layer

The cross-tool gap is exactly where a memory layer above the chat itself adds value.

A typical multi-tool day: Project-based morning work in Claude, afternoon brainstorming bounced between Claude and ChatGPT, evening coding in Claude Code. The Project knowledge base does its job inside Claude. The Artifact you built does its job inside Claude. But ChatGPT in the afternoon starts at zero context, and Claude Code in the evening reads CLAUDE.md not the Claude Project.

MemoryBase syncs conversations across ChatGPT, Claude, and Claude Code into one persistent store, so context built inside a Claude Project follows you to ChatGPT when you switch and to Claude Code when you open the terminal. Context Packs let you scope what loads into each session. The Claude-side workspace (Projects + Artifacts) keeps doing what it does. The cross-tool layer adds what Projects was never designed to handle: the rest of your AI surface. Letta and MemoryPlugin work the same category from different angles.

For the broader cross-tool context problem, see stop repeating context to AI. For where the AI coding tools split, see AI coding tools that actually help you ship in 2026.

Frequently asked questions

What's the difference between Claude Projects and Claude Artifacts?

Projects is a container for your work: knowledge base files, project instructions, and all the chats that share that context. Artifacts is the live output panel that opens to the right of your chat when Claude generates code, documents, visualizations, or interactive apps.¹² Projects is what you set up once. Artifacts is what you build inside any given conversation.

Can I use Projects on the free Claude plan?

Yes, up to 5 Projects on the free plan. Paid plans (Pro, Max, Team, Enterprise) lift that cap and add RAG over the knowledge base so larger collections stay performant.³

What file formats can I upload to a Claude Project knowledge base?

PDF, DOCX, CSV, TXT, HTML, ODT, RTF, and EPUB. Individual file size cap is 30MB. On free Claude, total knowledge content needs to fit in the context window. Paid plans automatically use RAG to handle larger knowledge bases.³

Can Claude Artifacts call external APIs?

Yes, on Pro, Max, Team, and Enterprise plans. Artifacts can make direct API calls and connect to MCP servers to read from and write to external tools like Asana, Google Calendar, and Slack.¹ Persistent storage (up to 20MB per artifact) lets stateful artifacts (journals, trackers, dashboards) keep data across sessions.

Should I upgrade from Pro to Max for Projects and Artifacts?

The feature set is the same. The difference is usage limits. Pro covers 1-2 focused sessions per day, Max 5x covers roughly 5x that workload, Max 20x covers 20x.⁴ If you're hitting Pro's session or weekly caps a couple of times in a typical week, Max 5x is the cheapest fix. If you're hitting Max 5x caps regularly, Max 20x is the next step before the API.

Sources

1. Anthropic Help Center, What are artifacts and how do I use them? Retrieved 2026-05-11.
2. Anthropic Help Center, What are projects? Retrieved 2026-05-11.
3. Anthropic Help Center, How can I create and manage projects? Retrieved 2026-05-11.
4. Anthropic Help Center, Use Claude Code with your Pro or Max plan. Retrieved 2026-05-11.
5. Anthropic Help Center, What is the Pro plan? Retrieved 2026-05-11.