AI Tooling Quickstart

Get started using Openfort's AI tooling to build your app faster than ever.

Steps

Install Openfort's MCP server.
Add rules for the LLMs.
Create a new project.
Debug common issues.
Discover all the capabilities.

1. Install Openfort's MCP server

This will allow your AI Assistant to interact with Openfort's tools on your behalf to create projects and manage them.

Ensure you have the following prerequisites:

Node.js - Installation guide here
An Openfort account - Create one here

Now, add it to your code editor. Based on your preferred tool, follow the instructions below:

To integrate our MCP Server with Cursor you can either:

One-click installation

Edit the `~/.cursor/mcp.json`

You can look it up on your system or find it under the Tools & Integrations tab in your Cursor Settings. Fill it with the following content:

{
  "mcpServers": {
    "openfort-mcp": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://mcp.openfort.io/sse"
      ]
    }
  }
}

Then you should see the Openfort MCP server listed on your Tools & Integrations tab without the need to restart. The authentication will trigger automatically.

For integration with Windsurf, replace the contents of the ~/.codeium/windsurf/mcp_config.json file with the following. It can be located at:

Windsurf Settings > Cascade > Plugins (MCP Servers) > View Raw Config

MacOS/Linux

{
  "mcpServers": {
    "openfort-mcp": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://mcp.openfort.io/sse"
      ],
      "disabled": false
    }
  }
}

To integrate an MCP Server into VS Code for use with GitHub Copilot, you should edit the .vscode/mcp.json file or run the MCP: Open User Configuration command which opens the file to add the following content:

{
  "servers": {
    "openfort-mcp": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://mcp.openfort.io/sse"
      ]
    }
  }
}

To add our MCP Server to Claude Desktop, click on Edit Config in the Developer tab under Settings to automatically create a file at:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json

Once created, fill it with the following content:

{
  "mcpServers": {
    "openfort-mcp": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://mcp.openfort.io/sse"
      ]
    }
  }
}

You will need to restart Claude Desktop after modifying the configuration file. Also, if you freshly installed Node.js you may need to reboot your computer too.

2. Add rules for the LLMs

LLMs tend to forget about the availability of tools and can make some pathological mistakes. Therefore, it's a good idea to include rules to remind them about this. Here's how to set them up.

3. Create a new project

Now create a new project in your editor and type up a prompt to the LLM to scaffold it for you. When needed, the LLM will automatically call the available tools on Openfort's MCP Server, enhancing your developer experience. Here's an example prompt you can try:

An Openfortkit App

4. Debug Common Issues

Here are some common issues you might encounter and how to resolve them.

General Errors

After the agent finishes creating a project, it may still throw errors, even with extended context. This is especially common for complex prompts or large applications.

Don’t expect the AI to flawlessly generate entire applications in a single prompt without any issues.

To resolve these errors, fix them manually or ask the AI for help. Iteration is normal, review the output, make corrections, and continue prompting as needed.

Loop When Creating a Policy

Occasionally, the AI agent may get stuck in a loop while creating a policy. The policy is successfully created, but the agent repeatedly attempts to update it with the same values.

The cause is unknown, and the effect is harmless.

To fix this, simply cancel the generation and prompt the agent to continue with the next step.

npm Error: Missing script: "dev"

If the AI agent fails to start the project using npm run dev, it’s often because it created the project in a subfolder and didn’t change into that directory before running the command.

Manually navigate to the subfolder and run the project again.

No Permission to Edit the .env File

When the AI agent fails to edit or create a .env file with your project keys, it’s usually due to insufficient file permissions.

In Cursor, add a .cursorignore file with !.env to explicitly allow the AI to edit the .env file.

For other editors or environments, follow an equivalent approach to ensure the file is not ignored.

5. Discover all the capabilities

For more information on the available tools and how to use them, check out the MCP Server documentation.