Teams using Claude and ChatGPT together: how to share work between them

The short answer

To share work between Claude and ChatGPT, pass the conclusion, not the chat. Each assistant’s sharing features work inside that assistant and its workspace, so a Claude link or a ChatGPT project does not reach a colleague on the other tool. Use a neutral format (summary, facts, files, deadline), or a connector that both assistants can use.

Many teams end up with both assistants without ever deciding to. Developers pick one, sales picks the other, someone brought a personal habit to work. That is a reasonable outcome. Forcing everyone onto one tool costs goodwill and rarely sticks. But it does create a boundary that work has to cross, and the built-in features were not designed for that.

What breaks at the boundary

Here is what happens when Anna, on Claude, needs to pass a client proposal to James, on ChatGPT.

Share links stay in their own world

Claude’s Share button creates a snapshot of the chat up to the moment of sharing. Attached files are not included, and on Team and Enterprise plans the link only opens for people in your organisation. James can read it if he has access. He cannot continue it in ChatGPT.

Projects are shared inside one product

ChatGPT shared projects let members with chat access see and use the project’s chats, files and instructions. On business plans, projects are shared only with members of your workspace. That is useful for a ChatGPT group, and invisible to Anna in Claude.

We compare the two kinds of project in more detail in ChatGPT shared projects vs Claude projects.

Files drift

Without a shared route, files travel by email or chat app. James downloads them, uploads them again to ChatGPT, and now there are two copies. When Anna fixes a number in hers, his is out of date.

Nobody knows who has it

An email saying “over to you” has no status. Anna does not know whether James read it. James is not sure whether he owns it or was only kept in the loop.

A practical setup for mixed teams

You do not need a big rollout. Five decisions cover most of it.

1. Agree a neutral handoff format. A short note with the same headings every time: next action and date, where it stands, facts, decisions, open questions, files, who to ask. It reads the same whichever assistant wrote it.
2. Keep final files in one place. The team drive or the project folder, not chat attachments. A handoff names the file and its version.
3. Pick one place where handoffs land. An email folder, a channel or a tool. The point is that people know where to look.
4. Ask for a confirmation. “Got it, it’s mine” closes the loop. Without it, two people think someone else has it.
5. Write down what may not be pasted into an assistant. Client personal data, passwords, anything under a confidentiality clause. The rule should be the same for both tools.

This works without any extra software. It does depend on everyone doing the copying and pasting carefully, every time.

Where a connector helps

Claude and ChatGPT can both be extended with connectors built on MCP, the Model Context Protocol. A connector gives the assistant a small set of tools that act in another system. If the same connector is added to both assistants, it becomes the shared route that their own sharing features do not give you. We explain the idea in plain English in what is an MCP connector, and what a team should check before enabling one in Claude connectors for teams.

How Colle works with both

Colle is a connector for exactly this boundary. Anna adds it to her Claude, James to his ChatGPT. Anna says:

Claude finds James in the company, writes a summary, records the facts and attaches the proposal. James gets a notification straight away, accepts in one click and asks ChatGPT “what did I get?”. ChatGPT has the summary, the facts and the file. When James finishes or sends it back with a reason, Anna sees the status in Claude.

A few practical details for mixed teams:

• Everyone keeps their own assistant subscription. Colle connects them; it does not replace either.
• Sending from ChatGPT needs ChatGPT Business or Enterprise. The person you send to can work in either assistant.
• Microsoft Copilot and Gemini are coming soon. Today Colle works with Claude and ChatGPT.
• The format is the same both ways. A handoff from ChatGPT arrives in Claude with the same summary, facts and files.

Setup details are on the Colle for Claude and Colle for ChatGPT pages, and the full list is on integrations.

When you do not need Colle

• Everyone uses the same assistant and works from one shared project. Then the project is your shared space.
• People only need to read each other’s reasoning. Share links do that.
• Work rarely crosses between the two groups. The five-step setup above, with a good note and a confirmation, is enough.

If you are still choosing how your team should pass AI work around, the guide seven ways to hand off work between AI assistants puts the options side by side. For costs, see pricing.