SONUB
EN English 한국어 日本語 中文 हिन्दी বাংলা اردو Español Português Français Deutsch Русский العربية Bahasa Indonesia Tiếng Việt ไทย Türkçe

How to enable notifications

Notifications are unavailable in this browser or on this address. Try a supported browser on the secure SONUB site.

If notifications are blocked, allow them in your browser settings.

  1. Open the site controls beside the address bar, or your browser’s site settings. Set Notifications for this site to Allow.
  2. In your device settings, allow notifications for your browser and check Do Not Disturb or Focus.
  3. Return to SONUB and reload the page. Once notifications are on, the bell disappears.

On iPhone or iPad, add SONUB to your Home Screen, open it there, then allow notifications.

Admin Log in Display theme Default theme Dark theme Pink theme Chat 1:1 Chat Group Chat Forum Notices News Suggestions Introductions Main menu Log out

Claude Projects

A project is a workspace for one goal: its instructions, knowledge files and chats stay together, so every chat starts from what you already gave it.

Learn Claude Code
What it is

One workspace for one goal

A student at a laptop in front of one glowing workspace that holds a pinned instruction card, a stack of syllabus documents and a column of chat bubbles, with a small robot assistant beside it

Projects are self-contained workspaces, each with its own chat history and knowledge base.

Projects work on every plan, including Free. A Free account can make up to 5 projects.

Open my projects on claude.ai (new window)

What is inside a project

Project

One workspace for one goal

  • Instructions

    How should answer in this project — a role, a tone, the rules to follow. They apply to every chat in the project.

  • Knowledge files

    Documents, text and you upload. Claude uses them as in every chat in the project.

  • Chats

    The conversations you start inside the project. Each chat begins with the instructions and the knowledge.

Instructions and knowledge reach every chat. What you say in one chat does not carry to the next chat — unless you add it to the knowledge.

Try it

Make your first project — a study project for one course

Example: a project for your biology course. The syllabus and your notes become the knowledge, and Claude becomes a tutor that knows your course.

  1. 1Create the project

    Open Projects on claude.ai and press the button. Give it a name such as "Biology — Fall term". Claude cannot see the name or description; they are only for you.

    Button on screen: + New Project

    Open Projects (new window)
  2. 2Add the syllabus as knowledge

    Press + in the knowledge area on the right of the project page and upload the syllabus, your lecture notes and the reading list.

    Button on screen: +

  3. 3Write the instructions

    Tell Claude how to help in this project, then press Save instructions. You can copy this example:

    Button on screen: Set project instructions

    You are my biology tutor. Explain in simple words, use examples from the syllabus, and end each answer with one short quiz question.

  4. 4Start a chat in the project

    Open the project and start a chat. Claude already knows the syllabus and your rules. Try:

    Make a short quiz on chapter 3 of the syllabus, then check my answers.

How knowledge grows

A small robot with a magnifying glass pulls only two glowing pages out of a full bookshelf and puts them into a small box on a desk
  1. Small knowledge: Claude reads it all within the
  2. Knowledge nears the limit
  3. Paid plans: turns on by itself and searches only the parts it needs

At first Claude reads all the project knowledge at once, inside its context window. On paid plans (Pro, Max, Team, Enterprise), when the knowledge nears that limit, Claude switches to RAG mode by itself: it uses a project knowledge search tool to pull out only the parts it needs. There is nothing to set up.

How much knowledge a project can hold
  • Context window only 1×
  • With RAG (paid plans) up to 10×

Official help: RAG expands capacity up to 10 times. A project using RAG shows an indicator.

Share with your team — Team and Enterprise

On Team and Enterprise plans you can share a project with people in your organization. Each person gets one of two roles.

  • Can view

    Can see the contents, knowledge and instructions and can chat in the project, but cannot edit it.

  • Can edit

    Can change the instructions and knowledge, manage members and remove other members.

  1. Share project
  2. Add a name or email
  3. Choose a role
  4. Share

A project can be private (only invited members) or public to your whole organization. Even in a public project, your chats stay private unless you share them.

New · beta

The new Projects: one conversation, many cloud threads

A new of projects began rolling out with . Here a project is one ongoing conversation, and Claude splits the work into threads that run in the . Anthropic announced it on 2026-09-17.

A project used to be a folder — now it is a conversation

  1. Before

    A project = a folder

    A place to keep files and chats together. You opened each chat yourself, one at a time.

  2. Now

    A project = one ongoing conversation with Claude

    Ask for things as they come up. Claude works on each one in its own thread and reports back as it goes.

Example conversation — from the official introduction screen
  1. You: Checkout feels slow today. Can you look into it?
  2. Claude: Found the cause. A deploy this morning slowed it down.
  3. You: Great, fix it and add a test so it does not happen again.
  4. 4 replies
  5. Claude: On it — I opened a thread for the fix.
  6. Fix checkoutOpening a PR

The thread does the work on its own and reports back to this conversation when it finishes. Open its card to see the details.

A developer at a laptop with one big conversation bubble where a small robot conductor sends glowing cables to four clouds, each carrying a busy robot working at the same time
  1. You

    Paste tasks, reports or a whole list

  2. Project conversation

    Claude coordinates and starts a thread for each task

Threads running in parallel (example tasks)

  • Fix a bug in the login page
  • Add tests for the payment code
  • Summarize a folder of support tickets
  • One conversation, many threads: the conversation branches into a thread only when the work needs it.

  • Set context once: every new thread starts with the project instructions, , and files — and what came before.

  • Each thread is usually a cloud session — Claude Code running in the cloud, not on your computer. Cloud threads keep going after you close your laptop, and you can check on them from your phone.

  • When a task needs something only your computer has, ask Claude to run that thread in a folder on your computer instead of the cloud. This needs Remote Control and runs only while the computer is awake.

  • A thread that changes code works on its own and can open a .

The Overview pane sorts threads by state, for example:

  • Working
  • Waiting on you
  • Ready for review
  • Resolved

Who can use it: it is a public beta on the Pro, Max and Ultra plans and is rolling out step by step. Team and Enterprise do not have it yet. If Projects is not in the sidebar of claude.ai/code (also check under More), you can join the waitlist.

It draws from your plan usage, and several threads at once use it faster. Check Thread model and Thread effort in Project settings, and start with a few threads.

Some existing Pro and Max projects are being moved to the new version. Nothing is deleted.

Classic projects and the new Projects
Classic project New Projects (beta)
Where Claude chat (claude.ai) Claude Code (claude.ai/code, desktop app, mobile app)
Who does the work You chat with Claude, one chat at a time Claude starts threads and runs them in parallel
What every chat or thread starts with Instructions and knowledge files Instructions, memory, repositories and files
Plans Every plan (Free: up to 5 projects) Pro, Max and Ultra (public beta, rolling out)
How to use · new Projects

Using the new Projects — hand off work, steer it in words, come back to check

You use the new Projects the way you would brief a chief of staff. Write the instructions first, set how Claude works by simply saying so, then step away and come back to review the results. Below is how the official docs say to use it, in order.

It works at claude.ai/code, in the desktop app and in the Claude mobile app. It does not work in the , the extension or the plugin.

1. Create a project, or use something else?

One test: does the goal outlast one session and keep producing tasks? Then it is a project.

Work that suits a project

  • One goal across many repositories

    Example: "Bring every service up to the new config." Each repository gets its own thread and pull request, and Overview shows which ones are ready for review.

  • An area you keep feeding

    Paste the bugs, and review requests for one service into the conversation as they reach you. After one fix, say "remember this" and the next thread starts knowing it.

  • A build or migration bigger than a session

    Example: "Build what docs/spec.md describes." The work splits into threads, and decisions you asked Claude to remember early on reach the later threads.

  • Work that is not code

    Material you keep asking new questions about, like a folder of contracts or a support-ticket export. Upload the files instead of adding a repository, and each write-up a thread makes lands as a file on the Library tab.

When something other than a project fits better
Situation Use instead
One task that fits in a session — e.g. "Fix the flaky login test." Start a cloud session yourself
Every task needs your machine — a , a device , an behind your VPN A local session ( view to run several at once)
One task that repeats on a schedule with no conversation — e.g. "Post a dependency report every Monday." A cloud scheduled task on its own
Several people giving Claude work and steering it together in a Slack channel Claude Tag

2. Before you send a lot of work — your first batch in four steps

  1. 1 Write project instructions

    Click the gear above the project → Project settings > Memory > Project instructions (up to 16,000 characters). Every new thread and the project conversation start from this text.

  2. 2 Test with one small task

    Send just one small piece of the real work. When it finishes, open the thread to see how it reported back and what it did on its branch.

  3. 3 Check model and effort

    Check Thread and Thread effort in Project settings > General. A new project runs every thread on Opus at high effort, which uses your plan fastest.

  4. 4 Propose first · a few at a time

    At first, say "Propose threads and wait for my go-ahead" and "Run only a few at a time". Drop these limits once threads come back the way you want.

Five things a good brief covers

  • What the project is for
  • Where the work happens — repositories, the branch to start from, how to name pull requests
  • How a thread checks its own work before calling it done
  • What to do when something it needs is missing
  • What needs your go-ahead first

Example brief — taken from the official docs. Copy it and adjust it for your project.

This project keeps p95 latency for the payments API under 200 ms, in the payments-api repository. - Branch from main and open one draft pull request per thread. - Before you call work done, run make test and make lint and paste the summary lines in your final message. - If you can't reach something you need, such as a repository, a secret, an API or a connector, say exactly what's missing in your first message and stop. Don't substitute, mock or guess. - Don't merge, force-push or change CI configuration without asking me in the thread.

Rules for one repository only, such as its build commands, belong in that repository's CLAUDE.md, which cloud threads read. When you correct a thread during the work, also say "remember this" — it goes into project memory and later threads start with it.

3. Set how Claude works by saying so

A quick question gets an answer right in the conversation. New work goes to a new thread or to a thread already working in that area, and Claude tells you which. Several unrelated tasks in one message become separate threads. To change how Claude works, say things like these in the conversation.

What you want and what to say
What you want Say this
Confirm before starting "Propose threads and wait for my go-ahead before starting them."
Start right away "Start these now without asking me to confirm."
Run fewer at once "Run at most two threads at a time."
Keep one area together "Reuse an existing thread for follow-ups in the same area."
Fewer updates "Only post when something finishes or is blocked." · "Post shorter updates."
See the whole picture "Give me a status update on every thread."
Save usage "Do this task with a smaller model."
Careful pull requests "Don't open a pull request until I've seen the plan."
Diagnosis only "Tell me what's wrong in these repositories and don't fix anything yet."
An answer without a thread "Answer that here instead of starting a thread."

Claude saves requests like these to project memory and follows them, but they are not enforced settings. Even "at most two" is not a hard cap. To word a rule exactly and apply it to every thread from the start, put it in the project instructions.

4. When you hand off work and step away

Cloud threads keep working after you close your laptop, and you can check on them and steer them from your phone. A thread on your own computer works only while that computer is awake. Take care of four things before you leave.

  • Say "go ahead without asking me"

    A thread that needs approval stops and waits. The approval prompt is inside that thread, so saying "go ahead" in the project conversation does not reach it — open the thread and answer there.

  • Say "commit and push as you go"

    If a cloud thread's workspace cannot be resumed, the thread continues from a fresh copy of the repository. Changes that were not committed can be lost.

  • Leave out work that needs your computer

    A thread on your computer stops when the computer sleeps. Send overnight work to cloud threads.

  • It continues after a limit

    When a thread hits your plan's usage limit it waits, then continues on its own when the limit resets (Auto-continue when usage limits reset in Project settings > General, on by default).

When you are back, look in Overview in this order

  1. Waiting on you Threads waiting for you
  2. Ready for review Pull requests to review
  3. Library Files the threads made

Limits and notifications to know

  • You can start up to 500 new threads a day across all your projects.

  • A thread with no activity for 7 days becomes Resolved on its own.

  • The desktop app notifies you when Claude posts in the conversation, a thread hits an error, or a thread needs your input. In a , check the dot on the Overview button.

5. Work that needs your computer — Work locally

Only when a task needs something just your computer has, such as a local database, a device emulator or an API behind your VPN, run that task's thread on your computer. The other threads keep working in the cloud.

  1. Claude Code v2.1.280 or later
  2. Connect the folder
  3. + menu → Work locally
  4. Allow once on the card

Connect with the desktop app: in Settings > Code, turn on Use this computer from your phone and claude.ai and add the folder to the list under it. Threads run while the app is open.

Connect with a terminal: run this command in the folder and leave it running. In a repository, --spawn gives each thread its own worktree.

claude remote-control --spawn worktree

A thread on your computer uses that computer's files, tools, , hooks and rules. It gets the project instructions but does not load the memory files.

6. Saving usage — models and effort

A project draws on the same plan limits as your other Claude Code sessions and uses them faster than one session (especially on Pro). Running threads, the project conversation that reads reports and decides what is next, and threads watching a pull request use your plan. An idle or archived project does not.

  • Lower the model or effort

    In Project settings > General, change Thread model and Thread effort for threads, and Coordinator model and Coordinator effort for the project conversation. Changes save at once and apply to new threads. To use a different model for one task, ask for it in that task.

  • New work in a new thread

    The lasts only 1 hour. A large thread that sat idle longer than that re-reads its whole conversation when a follow-up reaches it.

  • Run fewer threads at once

    Run fewer threads at a time, and let the conversation answer small questions instead of starting a thread.

  • Stop watching pull requests

    A thread watching a pull request wakes up and uses your plan when fails or a review comment arrives. Tell that thread "stop watching the pull request".

  • Check the Usage screen

    Project settings > Usage shows use by thread and by model, and how much went to the project conversation.

7. When you get stuck

Symptom and what to do
Symptom What to do
A thread is waiting for approval Open that thread and answer inside it.
Several threads guessed wrong or stalled Ask in the conversation: "For every open thread, list what you asked it to do, what it assumed or couldn't reach, and what it's waiting on." Close threads that started from a wrong assumption as Resolved, fix the gap in the instructions or environment once, then test with one thread first.
You hit the usage limit By default the work continues on its own when the limit resets. If you don't want that, turn off Auto-continue when usage limits reset.
You can't find the results Open the thread's card in the conversation, or look at the Library tab in Overview.

To study further: this part is based on the study collection claude-projects-study, checked against the official docs.

See the study notes on GitHub (new window)

Tips and cautions

  • Never upload secrets

    No passwords, or private keys. Knowledge is read in every chat, and on Team plans also by the people you share with.

  • The name is not an instruction

    Claude cannot see the project name or description. Put what Claude must know into the instructions or the knowledge.

  • Keep what matters in the knowledge

    If one chat produced a good summary, add it to the knowledge so later chats can use it.

  • Give Can edit carefully

    Members who can edit change the instructions and knowledge for everyone. Give it only to people who need it.

  • Watch your usage

    Large knowledge and many parallel threads use your plan faster. Keep only the files the goal needs.

  • Check against the source

    Claude can still make mistakes. Compare important answers with the syllabus or the textbook.

Official sources

Checked against the official help on 2026-11-09. Plans and screens change, so read the official pages too.

Glossary

Pick a word you don't know. You'll see a simple explanation and related terms.

No matching term yet.

Pick a term to see its explanation.

Reading window