MCP Server Integration

#

Tinkerwell includes a built-in MCP (Model Context Protocol) server that allows external AI tools to run PHP code in your Tinkerwell projects. This enables seamless integration with AI assistants to execute code snippets and access your project data.

💡 The MCP server does not need a configured API token because your other AI tools simply use it as a way to run code in the application context, even if the context requires connecting via SSH first. It's exposing Tinkerwell capabilities for machines, AI agents and everything that supports us in our work as developers in the future.

Supported AI Tools

#

Cursor

#

• One-click configuration through Tinkerwell settings
• Automatically configures MCP client connection
• Requires Cursor restart after configuration

Claude Desktop

#

• One-click configuration (macOS and Windows only)
• Automatically configures MCP client connection
• Requires Claude Desktop restart after configuration

Manual Configuration

#

• Copy JSON configuration for custom MCP clients
• Works with any MCP-compatible AI tool
• Includes server path and connection details

Setup

#

1. Open Tinkerwell Settings → Intelligence
2. Choose your preferred AI tool and click "Configure"
3. Restart your AI tool to apply the configuration

Available MCP Tools

#

The Tinkerwell MCP server exposes these tools for AI assistants:

Code Execution

#

Both execution tools run PHP within the context of your project, granting access to all its classes, functions, and frameworks (like Laravel, WordPress, etc.). Local execution works instantly. Remote execution, however, requires an active SSH connection—this is automatically established by the MCP server when you launch it. Keep in mind that this may perform actions on the remote system, so always review your code carefully before running it, as the results could be destructive.

Snippet Management

#

AI tools can both read existing snippets and create new ones. New snippets are automatically associated with your current project and support custom labels for easy organization.

Docker Environment

#

⚠️ The evaluate-local-php-code tool currently does not use Docker runtime configurations set in your Tinkerwell project settings. If your application runs in a Docker container, use this SSH-based workaround.

SSH Workaround for Docker

#

1. Setup SSH access in your Docker container by configuring SSH server and exposing port 22 in your docker-compose.yml

2. Create SSH connection in Tinkerwell pointing to your Docker container

3. Use evaluate-remote-php-code instead of evaluate-local-php-code when working with AI tools

The AI will execute code within your Docker container context, providing access to your application's full environment, database connections, and dependencies.

Example Usage

#

Once configured, your AI assistant can:

Run this PHP code in my Laravel project:
User::where('active', true)->count()

Save this code as a snippet:
collect([1, 2, 3])->map(fn($n) => $n * 2)

List my remote connections and run code on production