What ChatGPT shared projects give a team
A shared project in ChatGPT gathers one project’s chats, files, and custom instructions in a single place. People working on that project open the same folder and start from the same context, so a teammate does not have to re-explain the work in every new thread. For a team that lives inside ChatGPT, that is often enough.
ChatGPT also has a separate layer called company knowledge. It lets ChatGPT read connected company sources, but it is read-only, and it is limited to Business, Enterprise, and Edu plans. The assistant can pull that knowledge in, but it cannot write back to it.
Why teams look past shared projects
Two limits push teams to look further. First, the context stays inside ChatGPT. If a teammate opens Claude, Cursor, or Claude Code for a task, none of the shared project context follows them. Second, the parts an assistant can edit and the parts it can only read are split apart. Real project memory needs to be written by the assistant that just learned something, not only read.
Most people do not use one assistant all day. The right tool for the moment changes, so a chat happens here and a coding agent runs there. When project memory is tied to a single vendor, every switch means carrying the same decisions, file names, and “do not touch” notes by hand.
What to look for in a cross-assistant option
When you compare alternatives, check three things shared projects do not cover:
- Assistant-neutral. The same store opens from Claude, ChatGPT, Cursor, and Claude Code, not one walled app.
- Agent-writable. The assistant updates the shared notes directly, so the memory grows while the work happens.
- Yours to leave with. You can export everything as plain files, with a clear record of who changed what.
agent-wiki: one neutral place every assistant reads and writes
agent-wiki is a hosted markdown wiki that lives outside any single assistant. It speaks MCP, so Claude, Claude Code, Cursor, and ChatGPT (through its developer-mode connector) all read and write the same files. Your project context is not trapped in one vendor’s project folder. It is one source of truth that whichever assistant you open that day can use. Every read and write is logged with full version history, so you can see what an agent changed and revert in one click. You can export any wiki as plain markdown at any time.
See why it exists, or read how access stays private and revocable on the security page. For plans and the full feature list, start on the home page.
Connect your assistants to one shared wiki
Give every teammate’s assistant the same project memory, whatever tool they open that day. Connect Claude, ChatGPT, Cursor, or Claude Code to a shared agent-wiki and keep your context where no single vendor can lock it in.