For many people ChatGPT is the first AI of the working day, so the question comes up quickly: can it read the Merlin plan too? For direct access the answer is no, and the reason lies in the technology behind it. MCP servers come in two kinds, as a local program on your own machine and as a service with a web address. ChatGPT so far speaks only to the second kind, and the Merlin server is deliberately the first. Where that difference lies, why it protects your data, and which two routes stay open with ChatGPT is what follows.
The video for this article
The video (in German) is embedded in the German version of this article.
Why does ChatGPT of all things fail here?
ChatGPT reads calendars, analyzes spreadsheets, and searches entire mailboxes. It still cannot reach a project plan on your own computer. The difference lies in how an AI gets additional data at all. There are two kinds of MCP servers:
| Local server | Server on the internet | |
|---|---|---|
| Where does it run? | As a small program on your Mac | At a provider on the internet |
| How does the AI talk to it? | Via the program's standard input (stdio) | Via a web address (HTTPS) |
| Who can reach it from outside? | Nobody | Anyone with the address and credentials |
| Example | The Merlin server | The connectors ChatGPT requires |
The Merlin server is deliberately a program of the first kind. We built it so that nobody from outside can reach it. ChatGPT currently only supports the second kind: its connectors require a reachable address on the internet, and our server has none.
And if you put it on the internet?
Technically possible. A tunnel service can make a local program reachable under a public address. But then you abandon the promise we built the server for: from that moment, someone from outside can very much reach your Mac, and you carry the responsibility that it is the right people. We do not recommend this route and therefore do not describe it.
So your project plan cannot go online at all, because the server does not allow it. The same property that stops ChatGPT protects your data.
Route 1: the Intelligent Assistant in Merlin Project
ChatGPT is already inside Merlin Project, in the Intelligent Assistant. It builds a complete new project from a text description: you type three sentences, and Merlin creates phases, tasks, and milestones from them. This is built in and needs no MCP server. The article Creating projects with AI shows how it works in detail.
The difference from MCP access: the assistant creates new projects. It does not answer questions about a running plan.
Route 2: the copy bridge
If you want to keep thinking in ChatGPT, have a supported client such as Claude Desktop output the project data as a table, copy it, and paste it into ChatGPT. The prompt for that:
Prompt: How do I get my plan as a table into another tool?
Output all tasks of the active project as a Markdown table, grouped by phase: columns task, start, finish, duration, responsible resource, and progress in percent. Mark milestones in the task column with a leading diamond symbol. No explanations before or after, only the table.Open this prompt in the library ›
The result is a Markdown table that ChatGPT picks up cleanly. It is a snapshot, though, with the same limits as a PDF export: as soon as the plan changes, copy again. If you want to work with the plan permanently, choose a client that talks to the local server directly.
These clients talk to Merlin directly
- Claude Desktop: One block in one file, restart
- LM Studio: Download a model, one block in one file
- Claude Code: One command
- Codex: One command
- Antigravity: One block in one file, full restart
- ChatGPT (this article): Copy bridge or Intelligent Assistant
- Which AI client suits you?
All five clients use the same server and the same per-project sharing. Which one suits you is settled by the decision helper at the end of that list in a few questions.
If anything changes
As soon as ProjectWizards offers an optional remote mode for the MCP server, or ChatGPT supports local servers, we will update this article and the video. Until then: you decide who reads your plan. If your company ties you to ChatGPT, tell us below the video or in the forum. That feedback reaches the development team and helps decide how high remote access sits on the list.
The full list of clients with setup instructions is in the MCP documentation. To try the copy bridge, use the sample project Produktlaunch Phoenix.
If you have any questions about this blog article or would like to discuss it, we look forward to your contribution in our forum.
