Lovable MCP server

1. 1

   Check your Lovable account plan and workspace settings to ensure third-party MCP clients are enabled.

   This is a prerequisite for connecting your AI client to Lovable, especially for Enterprise workspaces where it might be disabled by default.

2. 2

   Choose your AI client (e.g., ChatGPT, Claude, Cursor, VS Code) and follow its specific instructions to add the Lovable MCP server.

   Each AI client has a slightly different method for connecting to an MCP server, so following the correct steps ensures a successful connection.

3. 3

   Authenticate with your Lovable account when prompted by your AI client.

   This step uses OAuth to securely link your AI client to your Lovable account, granting it the necessary permissions to manage your projects.

4. 4

   Start using natural language prompts in your AI client to create, iterate on, inspect, or deploy Lovable projects.

   Once connected, your AI client can use Lovable's tools to automate development tasks, allowing you to build and manage apps efficiently.

Connect Lovable to my AI client using the Model Context Protocol. I want to be able to create, edit, and deploy Lovable projects directly from my AI chat. Provide the necessary URL and authentication steps for my chosen client.

Model Context Protocol (MCP)

An open standard that allows AI tools to discover and use other software applications, like Lovable, by understanding what they can do.

OAuth

A secure way for one service (like your AI client) to get permission to access your information or act on your behalf in another service (like Lovable) without sharing your password.

AI client

A software application, like ChatGPT or Claude, that you interact with using natural language to perform tasks, often by connecting to other tools.

Programmatically

Controlling or managing something using code or automated instructions, rather than manual clicks or direct human interaction.

• Connecting a client grants it the same access you have to all your Lovable projects, so be mindful of the client's capabilities and trustworthiness.
• Tool calls made through the MCP server use real Lovable credits and make live changes to your projects.
• Deploying a project (`deploy_project`) makes it live and publicly accessible on Free and Pro plans, or according to your workspace's default settings on Business and Enterprise plans.
• Running SQL queries (`query_database`) through the MCP server uses your full database permissions, including read, write, and schema changes, which can alter your data.
• If your workspace has Enforce SSO enabled, you may need to re-authenticate via your browser when your session expires.