What are Claude Projects, and which version do you have?
A Claude project is a self-contained workspace with its own chat history, its own knowledge files and its own instructions. You upload the brief once, write the rules once, and every chat inside the project starts from them. Anthropic's help center says projects are available to all users, free accounts included, and that free users can create a maximum of five projects.

There are now two Claude products with this name. Anthropic's Claude Code docs open the second one with a plain definition: "A project is one ongoing conversation where Claude coordinates a stream of related work for you." Each task becomes a thread, usually a cloud session. The help center page for the classic version now says a new version is rolling out in stages, and that some existing Pro, Max and Ultra projects are being moved to it. Anthropic emails people first if they used a project in the last 30 days.
| Classic Projects | Projects beta | |
|---|---|---|
| What it is | Chats, files and instructions in one workspace | One coordinating conversation that starts parallel threads |
| Plans | All, including Free (5 projects) | Pro, Max and Ultra, gradual rollout; not Team or Enterprise |
| Where | claude.ai, the Projects page | claude.ai/code, the Code tab of the desktop app, the mobile app |
| Instructions | Set per project, no published character cap | Up to 16,000 characters |
| Files | 30 MB each, total must fit the context window | Up to 1,000 files and 2 GB per upload, 480 MB for one file |
| Sharing | Team and Enterprise only | None, a project belongs to one user |
| Memory | Per project, can be separated from account memory | Project memory files Claude writes itself |
You can tell which one you have in a minute. Open claude.ai/code and look for Projects in the sidebar, or under More. If it's missing in both places, the beta hasn't reached your account, and Anthropic's docs point to a waitlist form. Everything in the how-to section below works either way, and the beta has its own steps right after the classic ones.
How do you create a project in Claude?
You create a classic project from the Projects page in about a minute. These steps come from Anthropic's help center page on creating and managing projects, read on 15 November 2026.
- Open Projects in the left panel on claude.ai, or go to claude.ai/projects.
- Select + New Project in the upper right.
- Type a name and a description. Claude can't see these two fields, so they're labels for you.
- On Team or Enterprise, choose whether the project stays private or is shared with your organization.
- Use the + button on the right of the project page to upload documents, text files or code snippets. Every chat in the project can use them.
- Select Set project instructions, describe how Claude should answer, and save. They apply to every chat in the project.
Memory has its own switch. Each project keeps separate memory, and you turn off Use account memory under Memory on the project page to keep it fully apart from your account. To skip memory for a single chat, turn Memory off in the + menu before you send the first message. Context doesn't travel between chats unless it sits in the project knowledge.
The beta has a different flow. Select Projects at claude.ai/code (or under More), then New project. Only the name is required. You can add one line under What do you want to work on?, then the GitHub repositories, files, folders or Google Drive folders threads should read. Click Create project. Standing rules go in Project settings > Memory > Project instructions, opened from the three-dot menu in the project header. A new beta project runs every thread on Opus at high effort, so Anthropic's docs suggest checking Thread model and Thread effort before the first batch.
To retire a classic project, open the three-dot menu on the project page and confirm the archive. Archived projects keep their members, permissions and knowledge. You can't delete one until you unarchive it, and Delete then asks for a second confirmation. To delete, use the three-dot menu on the Projects page or inside the project, select Delete and confirm with Yes, delete.
What are the limits of Claude Projects on Free, Pro and Max?
Free gets up to 5 projects, Anthropic states no project cap for Pro or Max, and the file rules are the same on every plan. Claude's pricing page, read on 15 November 2026, lists Free at $0 with up to 5 projects, Pro at $20 a month or $17 a month on the annual plan, and Max at $100 a month, both with projects included. Anthropic publishes no number for how many projects Pro holds.
Files are limited by size and by context, not by count. A project file can be up to 30 MB, and Anthropic says the number of files is unlimited as long as the content fits in Claude's context window. Project files are read as text, except PDFs that carry charts or images. A chat upload is a different budget: 500 MB a file and 20 files per chat. The details are on the Claude file upload limit page.
Paid plans get one extra. When project knowledge nears the context limit, Claude switches the project to retrieval mode on its own and fetches only the passages a question needs. Anthropic says this expands capacity by up to 10x on Pro, Max, Ultra, Team and Enterprise. Free accounts don't get it, so a Free project that outgrows the context window is a project you must trim. A visual indicator shows retrieval is on, and Claude then runs a project knowledge search when it reads your files.
The beta counts differently. It doesn't cap files so much as usage: Anthropic says a project draws on the same plan limits as your other Claude Code sessions and uses them faster, so on a Pro plan, watch your usage on days you run one. The enforced ceiling is 500 new threads a day across your projects. Each cloud thread has its own context window, and the docs say you don't manage context windows in a project because threads compact automatically. Claude Free vs Pro has the plan comparison around these rows.
What is a Claude project good for?
A project pays off when you return to the same material and the same rules more than three times. Below are five uses that fit, with the instruction line we would paste. These are our suggestions, not Anthropic's, and we haven't run them as a benchmark.
- A client or course folder. Pin the brief and the style guide. Instruction: "Write for a first-year nursing student, in plain English, and quote the pinned handbook when you give a rule."
- A contract or policy library. Upload the PDFs once and ask new questions each week. On a paid plan, retrieval lets the library grow past one context window.
- A writing voice. Pin three of your own posts and tell Claude to match their sentence length. Each new chat starts with your voice, not a blank page.
- A research notebook. Keep papers and notes in one project, and ask Claude to say which file a claim came from.
- A stream of engineering tasks (beta). Paste a bug report or a stack trace, and Claude starts a thread for it. Anthropic's examples include one goal across many repositories, where each thread opens its own pull request.
Two habits make any of these work. Scope a project to one stream of work, because one catch-all project dilutes the instructions. And keep the knowledge current: replace a stale draft instead of adding the new one beside it, since both get read.
What can a Claude project not do?
A classic project can't share context between chats, can't read images in pinned files, and can't be shared outside Team and Enterprise. Each chat is its own conversation. Anthropic's help page is blunt that context isn't shared between chats unless it sits in project knowledge. Sharing, with Can view and Can edit permissions, is a Team and Enterprise feature, so a Pro user can't hand a project to a colleague.
The beta has its own walls, all from Anthropic's docs on 15 November 2026. "Projects aren't offered on Team or Enterprise plans yet." A project belongs to one user, and you can't share it or its threads. A thread belongs to the one project that started it and can't move to another. Projects don't exist in the terminal CLI, the VS Code extension or the JetBrains plugin, and they don't run through Amazon Bedrock, Google Cloud or Microsoft Foundry. A cloud thread's sandbox pauses between turns, and if it can't resume, the thread restarts from a fresh clone, so uncommitted changes can be lost.
Neither version swaps the model for you inside a chat from another vendor. Every answer in a Claude project comes from a Claude model. If you want the same folder to feed GPT or Gemini as well, that is a job for a different tool, which the last section covers.
How do Claude Projects compare with ChatGPT Projects and Gemini Gems?
Claude Projects has no file count, ChatGPT Projects has the tightest published caps by plan, and Gemini Gems is being retired on personal accounts. The table is compiled from each vendor's help pages on 15 November 2026, for individual plans.
| Claude Projects | ChatGPT Projects | Gemini Gems | |
|---|---|---|---|
| Free tier | Yes, up to 5 projects | Yes, 5 files per project | Personal accounts, ends November 2026 |
| Files | 30 MB each, total fits context; retrieval on paid | 5 on Free, 25 on Plus and Go, 40 on Pro, Business and Enterprise; 10 per upload | Optional Knowledge files, no number stated on the Gems page |
| Instructions | Set per project | Project settings, override your global custom instructions | Name and instructions per Gem |
| Memory | Per project, can be separated | Default or project-only memory | Not stated |
| Sharing | Team and Enterprise | Invites or links on every plan, Free included | Gems and chats can be shared |
| Models inside | Claude only | OpenAI only | Gemini only |
Google's own Gems page carries a notice that matters if you built your workflow there: "Starting in November 2026, Gems will go away on personal Google accounts and automatically transition to skills." Google's page also says Gems can't be used with Gemini Live. ChatGPT's help page says project sharing is available on Free, Go, Plus, Pro, Business, Enterprise and Edu, with collaborator caps of 5 on Free, 10 on Plus and Go, and 100 on Pro. Claude's advantage is the file rule, since it never asks you to prune to a count. Its disadvantage is that sharing sits behind a Team plan. For a closer look at the ChatGPT side, read ChatGPT Projects alternative.
Pick Claude Projects when your folder is a few large documents and you want long, careful reading. Pick ChatGPT Projects when you need people to collaborate on a project from a personal plan. Pick neither when you want one folder that several vendors' models can use.
What if the folder should work with more than one model?
That's the gap a vendor project can't close: the folder, the files and the rules belong to one company's models. Whizi Projects do the same three jobs with a different boundary. A project holds up to 10 pinned files of 10 MB each, plus instructions of up to 32,000 characters, and it's open to every signed-in account on every plan, free included. On every message the instructions and the pinned file text are folded into the prompt, whichever model answers.
The free plan gives 10 messages a day on Whizi AI, Auto and Claude Haiku 5.5, with no card. Pro carries Claude, GPT and Gemini in one app, and its first week is $1.99. So a pinned brief can be read by Claude in the morning and by GPT in the afternoon, with nothing to re-upload.
Whizi loses on three counts. It has no coordinating conversation that starts parallel threads, so the beta's repository work has no equivalent. Pinned files are text only, and images can't be pinned. And Whizi doesn't run Claude's own Projects or its retrieval mode: it's a separate feature with its own caps. The Whizi projects guide lists every limit.
- Check which version you have: open claude.ai/code and look for Projects in the sidebar or under More, since the beta is gradual
- On Free, plan around 5 projects and trim files by hand, because retrieval mode is paid only
- Write instructions about the reader and the format, then test them on one real task before you add more files
- Replace stale files in the project instead of adding the new version beside the old one
- In the beta, set Thread model and Thread effort before the first batch, since a new project runs Opus at high effort
- Turn off Use account memory on the project page if the project must stay separate from your other chats
Frequently asked questions
Are Claude Projects free?
Yes, in the classic version. Anthropic's help center says projects are available to all users including free accounts, and free users can create a maximum of five projects. Retrieval mode for large knowledge is paid only. The new Projects beta needs Pro, Max or Ultra.
What is the difference between a Claude project and a regular chat?
A project keeps knowledge files and instructions that every chat inside it starts from, and a regular chat starts empty. Chats in a project don't share context with each other unless it sits in the project knowledge. In the beta, the project is one coordinating conversation that starts threads.
Can you share a Claude project with someone?
Only on Team and Enterprise, with Can view or Can edit permissions. The beta can't be shared at all: Anthropic's docs say a project belongs to one user.
How do you get on the Claude Projects beta?
Check claude.ai/code for Projects in the sidebar or under More. If it is missing there too, the rollout has not reached your account, and Anthropic's docs say Pro, Max and Ultra users can join a waitlist. Team and Enterprise plans are not offered the beta yet.