Projects

Projects are sandboxed workspaces—each with its own chats, memories, and files. They let you keep work, clients, hobbies, or research in separate buckets so context stays clean and nothing leaks across boundaries you didn't intend.

Your main account has global memory and global files for everyday chats outside any project. When you open a project and start chatting inside it, the system uses only that project's memory and files. Global context does not leak in. Other projects stay isolated too. That isolation is intentional: you can have a project for a client, another for personal research, and another for a hobby, and each one keeps its own context.

Projects have their own chat list, so you can archive or remove project threads without touching your main list. Switching into a project loads its context instantly—icon, color, name, and all the memories and files you've added there.

Create a project from the chat sidebar: click the project icon or use the "+" control to add a new one. Give it a name, optionally a description and goal, and choose an icon and color. The project appears in your sidebar and you can start chatting inside it immediately.

You can rename a project, update its description or goal, and change its icon or color at any time. Each project has its own chat list; new chats started inside the project are scoped to that project by default. You can archive or delete project chats without affecting your global chat list.

Projects are stored in your account and sync across devices. Memory and file settings are per-project. The main project-memory toggle is the umbrella switch for that workspace. If it is on, Adaptive Memory: Read and Adaptive Memory: Write can be enabled underneath it as separate foreground tools for that same project.

Project scope is strict. When you chat inside a project, the system uses only that project's memories and files. Your global memories and global files are not injected into project chats. Likewise, project memories and project files never appear in global chats or in other projects.

This keeps "work brain" separate from "life brain." If you have a project for a client with sensitive context, that context stays in that project. If you have a project for creative writing with its own tone and references, those stay isolated too. Personas can travel across both—attaching a persona to a project chat applies that persona's style and keeps persona memory available through persona-scoped tools without merging the underlying project or global stores.

When foreground read/write tools are enabled in a project, smarter models can inspect much more of that project's local context, similar to agent-style workflows. That usually produces better grounded answers, but it can also increase usage because the model may take extra tool steps.

Project limits depend on your plan. Explorer (free) allows 1 project. Builder allows 5, Studio 10, Aurora 25, and Pro 100. File uploads per project are capped separately from your total library allowance: Explorer has no file uploads, Builder allows 5 files per project, and higher paid tiers can attach up to 10 files per project while still staying within their total file allowance. Max file size per upload scales with tier as well.

If you hit your project limit, you'll need to archive or delete an existing project before creating a new one. The same applies to files inside a project. Check Account → Usage or the upgrade prompt in the app for your current limits.