The Model Context Protocol (MCP) is an open standard designed to unify and secure how artificial intelligence accesses external data and tools.

In concrete terms, the Poool MCP allows your AI agent (Claude, ChatGPT, Cursor...) to interact directly with the Dashboard. You can view, create, or modify your Access appearances and Engage elements in natural language, without having to use the interface.

The MCP is currently supported natively by most general-purpose AIs (Claude, ChatGPT, Gemini, Perplexity, Cursor...), usually requiring a paid plan or activation by your IT administrator. Access conditions vary and evolve rapidly; please check with your AI tool.

The MCP provided by Poool offers actionable tools for Poool Access and Poool Engage.

How to connect to the Poool MCP

Configuration in the Dashboard

Generating the access token

Connecting to the Poool MCP requires a personal access token (PAT), generated from your Dashboard.

• Go to your Dashboard, in the [location to be confirmed based on release] section.
• Generate a new token.
• Select the apps you want to grant your AI agent access to.
• Copy the generated token: it will only be displayed once.

A token only grants access to the applications selected at the time of its creation. To add or remove an app, generate a new token.

MCP server URL

Here is the url to use:

https://api.poool.fr/api/v4/mcp

Connection based on your AI client

The Poool MCP is compatible with any client supporting the MCP standard (ChatGPT in developer mode, Gemini, Perplexity...). Connection steps vary depending on the tool; refer to your AI client's documentation to add a custom MCP connector using the URL and token above.
All elements created via AI are linked to the account used for the MCP connection. They will therefore appear in the asset creation/modification history.

Controlling rights

When creating a token, you can precisely choose its permissions, which applications it can act on, and the associated rights.
For example, you can limit the token to pre-prod or development apps to prevent the AI from making changes in production.
It is also possible to choose between read-only access or allowing modifications.

Example of use

Available actions

Your AI agent can perform the following actions via the Poool MCP:

Apps

• findApp: search for an application by its name or appID
• getApp: retrieve an application using its appID
• listApps: list the applications accessible by the authenticated user

Appearances (Access)

• createAppearance: create an appearance
• findAppearance: search for an appearance by its name
• getAppearance: retrieve a complete Access appearance
• updateAppearance: modify an appearance including all fields (content, CSS, palettes)
• deleteAppearance: delete an Access appearance from the database

Elements (Engage)

• createElement: create a new Engage Element, consisting of the appearance (content and styles) and display conditions (device, date, behavior, combinable with AND/OR)
• findElement: search for an Engage element
• getElement: retrieve a complete Engage element
• updateElement: update an existing Engage Element, including all fields (content, conditions, styles...).
• deleteElement: permanently delete an Engage element from the database

Skills

• listSkills: list the available design editor reference documents (blocks, styles, responsive, display conditions, Access/Engage instructions)
• getSkill: read specific documentation before generating a configuration

This list evolves with the product; your AI agent can list the available actions itself at any time.

Updated on: 25/09/2026