Chats are great until the third one about the same topic. You paste the same background, re-explain the same constraints, and the assistant answers as if it just met you.
Projects fix that. Create a project, drop your chats into it, and everything inside shares one context.
What a project carries
- Project instructions — tone, constraints, definitions written once, applied to every chat in the project.
- Shared memory — decisions and facts surfaced in one chat are available to the next one.
- A clean switchboard — the sidebar groups project chats together, so "where was that answer" stops being a search problem.
Why not just one long chat?
Because long chats decay. Context windows fill up, old details get compressed, and unrelated threads bleed into each other. A project keeps each conversation focused while the durable facts live at the project level.
Works with every model
Project context is applied before routing, so it does not matter whether your prompt lands on OpenAI, Anthropic, Google, or any other provider in the catalog — the model receives the same project briefing.
Open the sidebar, create a project, and move two related chats into it. The first time a chat "remembers" a decision you made in a different one, you will not go back.