Using ChatGPT, Claude and Gemini on your project plans with MCP

Asking an AI assistant “is the project late?” after pasting a list of tasks produces plausible but unreliable answers: the model knows nothing about dependencies, calendars or float, and the dates it suggests are not calculated by anyone.

The Model Context Protocol (MCP) changes that. The assistant does not guess: it queries the plan through tools that use the Planitude scheduling engine, and every change it proposes is recalculated with the same rules as the Gantt chart. This guide explains what MCP is, what the assistant can do with your projects and how to connect ChatGPT, Claude and Gemini.

What MCP is#

The Model Context Protocol is an open standard that lets AI applications such as Claude or ChatGPT connect to external data sources, tools and workflows; its documentation compares it to a USB-C port for AI applications^[1]. A service publishes an MCP server with a list of tools; the assistant discovers them and uses them when they help answer a question.

For project plans, this means the assistant does not work on a copy pasted into the chat, but on the real plan, with the permissions of the person who connected it.

What the assistant can do with your plan#

The Planitude MCP server provides reading and analysis tools:

• Project status: the list of projects and an overview that classifies each one as late, at risk, on track or completed, with the reason and the deadlines in the next seven days.
• Plan details: tasks, dates, dependencies and people, with the option to show only critical tasks.
• Critical path: the chain of tasks with no float, the near-critical ones and the distribution of float.
• Delay impact: “if demolition is three days late, how far does the finish date move?”, compared with the available float.
• Monte Carlo simulation: P50, P80 and P95 finish dates, the probability of meeting the plan or a deadline, and risk drivers.
• Workload: overallocated people and the tasks that cause the overallocation.
• Baselines, versions and templates of the plan.

Questions such as “which projects are at risk this week, and why?” or “how likely are we to deliver by December 15?” get answers based on calculations, not on the model’s guesses.

Edits with a preview#

With edit permission, the assistant can also create projects (empty, from a template or from a structured plan with phases, tasks and links) and change them: add tasks, update durations and progress, link dependencies, save a baseline, import an .mpp file.

Larger changes go through a preview: the assistant prepares a set of edits, Planitude returns the calculated result (new finish date, deadlines met or missed, conflicts, overloaded people, critical path) without saving anything, and only applies the edits after confirmation. If someone has changed the project in the meantime, the edits are rejected and the assistant has to read the plan again. ChatGPT also asks for confirmation before every write action by default^[3].

Security: OAuth, permissions and revocation#

Web apps such as ChatGPT, Claude.ai and Gemini connect with OAuth, the authorization mechanism defined by the MCP specification, based on OAuth 2.1 with PKCE^[2]. There is no password to share and no key to copy: the assistant opens Planitude, you sign in and see a consent screen that lists what it will be able to do.

• Reading is always included; changes are granted only by turning on Also allow changes.
• You can limit access to a single workspace.
• The assistant acts with your permissions: it only sees the projects you can see.
• It cannot delete projects or tasks or manage your account and keys.
• You can disconnect any app at any time from Profile › API keys › Connected apps.

For tools that prefer it, you can create an API key with read-only or read-and-write permissions and an optional expiry; the key is shown only once and revocation takes effect immediately.

Connecting ChatGPT#

Connecting custom MCP servers goes through developer mode, available on the web for Plus, Pro, Business, Enterprise and Education plans^[3]:

1. in ChatGPT, open Settings › Security and login and turn on developer mode^[3];
2. among the apps, create a new app, paste the URL https://planitude.app/mcp and choose OAuth as authentication;
3. sign in to Planitude and allow access;
4. in a chat, choose developer mode from the + menu and turn on Planitude.

In organizations, an administrator can restrict developer mode. We have also submitted the Planitude app to the ChatGPT directory, where it is currently under review by OpenAI.

Connecting Claude#

On Claude you add a custom connector, available on the Free, Pro, Max, Team and Enterprise plans; the Free plan allows one^[4]:

1. open Customize › Connectors and choose Add custom connector;
2. paste the server URL; no OAuth client ID or secret is needed;
3. choose Connect, sign in to Planitude and allow access.

On Team and Enterprise plans, an Owner adds the connector from Organization settings › Connectors, and each member then connects with their own account^[4]. The connector works in claude.ai, Claude Desktop and the mobile app.

Connecting Gemini#

In the Gemini app, MCP-based custom apps are currently limited by Google to personal accounts of adults in the United States, with Gemini Apps activity turned on; work and school accounts are not supported^[5]. If you meet these requirements:

1. on gemini.google.com, open Settings › Connected Apps and choose Add a custom app^[5];
2. paste the server URL and follow the instructions;
3. sign in to Planitude and allow access.

The connection is created on the web but also works in the mobile app^[5]. For every other account, Gemini CLI remains available, as described in the next section.

Developer tools: Claude Code, Codex, Cursor and VS Code#

Developer assistants accept either OAuth or an API key:

• Claude Code: claude mcp add --transport http planitude https://planitude.app/mcp; on first use, Planitude opens for consent^[6].
• Codex: a [mcp_servers.planitude] block in ~/.codex/config.toml, which applies to the CLI, the app and the IDE extension.
• Gemini CLI: gemini mcp add --transport http planitude https://planitude.app/mcp.
• VS Code with GitHub Copilot: add the server in .vscode/mcp.json and the tools appear in Agent mode^[7].
• Cursor: add the server in ~/.cursor/mcp.json or in the project file^[8].

Ready-made examples, with the address and a key placeholder, are in Planitude under Profile › API keys. The other integrations, from Jira to Teams, are on the Integrations page and in the guide to integrations; for the API and webhooks, see the Developers page.

Frequently asked questions#