How to connect ChatGPT to your Gantt chart with MCP

ChatGPT is good at turning a vague goal into a list of steps. The problem is what happens next: you copy that list into a spreadsheet or a planning tool by hand, and the moment a date moves, the plan and the conversation drift apart.

Connecting ChatGPT to ProGantt Flow removes that copy-paste step. Through MCP, ChatGPT works directly on your Gantt charts: it can create a project with tasks and dependencies, push dates back when something slips, assign people and tell you who is overloaded. Every change it makes shows up in the app straight away, for you and for everyone you share the project with.

This guide covers ChatGPT on the web and Codex, which includes the ChatGPT desktop app, the IDE extension and the CLI.

What MCP is, in one paragraph

MCP (Model Context Protocol) is an open standard that lets an AI assistant use tools from other apps. ProGantt Flow runs an MCP server at https://mcp.progantt.com/mcp. When you connect it, the assistant gets a set of tools such as "list projects", "add task", "move task" or "get resource load", and it decides when to use them based on what you ask. If you want the longer version, read what an MCP server for project management is.

Before you start

• A ProGantt Flow account. You can create one for free at app.progantt.com with Google or Microsoft.
• ChatGPT with access to custom MCP servers, or Codex. Whether you can add custom MCP servers in ChatGPT depends on your plan, and on Business and Enterprise workspaces also on what your admin has enabled.

The connector is included on every ProGantt plan, Free and Pro. Connecting through sign-in, as described below, does not use one of your access tokens.

Connect ChatGPT (web)

ChatGPT adds remote MCP servers as plugins. You sign in with your ProGantt account, so there is no token to copy or keep safe.

1. In ChatGPT, open Plugins (chatgpt.com/plugins), click + and choose Add custom MCP server.
2. Give it a name, for example ProGantt, and a short description.
3. Under Connection, paste the server URL, including the /mcp path: https://mcp.progantt.com/mcp
4. Choose OAuth as the authentication method. Read the warning, click I understand and want to continue, then Create as a plugin.
5. Install the plugin. A ProGantt page opens in your browser: sign in with the same Google or Microsoft account you use in ProGantt and click Allow.
6. Open a new chat, type @, pick ProGantt and ask: "List my ProGantt projects."

If ChatGPT lists your projects, you are connected.

Connect Codex

The Codex CLI, the Codex IDE extension and the ChatGPT desktop app share one MCP configuration, so you only need to add ProGantt once. Codex also signs in with your ProGantt account.

In the ChatGPT desktop app or the IDE extension:

1. Open Settings → MCP servers and click Add server.
2. Name it progantt, choose Streamable HTTP and enter https://mcp.progantt.com/mcp.
3. Leave the bearer token and header fields empty. Save and restart.
4. Click Authenticate next to the server and sign in with your ProGantt account.

Ignore the STDIO option. It is for servers that run on your own computer, and asks for a command, arguments and environment variables that do not apply here. ProGantt is a remote server, so it always uses Streamable HTTP.

In the Codex CLI, add this to ~/.codex/config.toml:

[mcp_servers.progantt]
url = "https://mcp.progantt.com/mcp"

Then sign in once with codex mcp login progantt, and type /mcp in Codex to check that the server is active.

Prefer an access token over signing in? The connector docs show how to pass one through an environment variable. The detail that trips people up is that Codex's bearer token field expects the name of the variable, not the token itself.

What to ask once it is connected

Talk to it the way you would brief a project manager. A few prompts that work well:

• "Create a ProGantt project called 'Website relaunch' with discovery, design, build and launch phases, realistic durations and dependencies between them." ChatGPT creates the project with phases as groups, subtasks inside them and the order expressed as dependencies.
• "In 'Website relaunch', design is running a week late. Push it and everything that depends on it back by 7 days, and tell me the new launch date." It reads the dependency graph, moves the affected tasks and reports the result.
• "Who is overloaded in 'Mobile app v2' over the next two weeks? Suggest how to rebalance the work." It checks the load per person for that window and proposes reassignments.
• "Mark 'API integration' as 60% done, tag it as backend and assign it to Maria." It finds the task by its title and applies the changes in one update.
• "What is still blocking the 'Beta release' milestone?" It lists the unfinished tasks the milestone depends on.

Refer to projects and tasks by their titles. The assistant looks them up for you; you never need an internal ID.

If you are new to building plans this way, how to create a project schedule covers the planning side, and an AI agent for project management explains how far you can delegate.

What ChatGPT can and cannot see

A connection is account-wide. Once connected, the assistant can see and edit every project you can, not just one. The usual project permissions still apply:

• Projects you own or that are shared with you as an editor can be changed.
• Projects shared with you as a viewer are read-only. If you ask for a change, the assistant gets an error and tells you so.
• Projects you have no access to do not exist as far as the assistant is concerned.

Each tool is marked as read-only or as a write, so a client can ask for confirmation before changing anything, depending on its settings. You can disconnect at any time from the client, and access tokens can be revoked from Connect AI in ProGantt.

Troubleshooting

It connects but finds no projects. You probably signed in with a different account from the one you use in ProGantt, for example Google instead of Microsoft. Disconnect, connect again and pick the right account.

"Couldn't connect" when adding the server. Check that the URL ends in /mcp. The bare domain does not serve the MCP endpoint.

You can't find the option to add a custom MCP server in ChatGPT. It depends on your plan and workspace settings. In a Business or Enterprise workspace, ask your admin. Codex is an alternative that works the same way with ProGantt.

Codex asks you to sign in again. Sessions renew automatically, but after a long period without use you will be asked to authenticate again. Run codex mcp login progantt or click Authenticate.

The full reference, including every tool and the setup for Claude, Cursor and VS Code, is in the ProGantt MCP documentation. If you use those editors, see managing a project plan from Cursor or VS Code.