Buzzy MCP Implementation
Buzzy has implemented the Model Context Protocol to allow any Buzzy app to be connected to AI assistants like ChatGPT. This means users can interact with their Buzzy apps through natural language conversations, asking the AI to query data, create records, and perform actions within the app.
Video Overview
Watch this demonstration of building and connecting a Buzzy app with MCP:
How Buzzy Implements MCP
When you enable MCP for a Buzzy app, Buzzy automatically generates an MCP server that exposes your app's functionality to AI assistants. This includes tools for interacting with your datatables, executing Buzzy functions, and displaying widget screens.
Automatic Tool Generation
Buzzy automatically creates MCP tools based on your app's data model. For each datatable in your app, Buzzy generates tools for common operations:
List tools allow the AI to retrieve records from a datatable. The AI can use query parameters to filter, sort, and paginate results. For example, a "list_recipes" tool would let the AI fetch all recipes or search for specific ones.
Get tools retrieve a single record by its ID. This is useful when the AI needs detailed information about a specific item.
Create tools allow the AI to add new records to a datatable. The tool's input schema is automatically generated from your datatable's field definitions.
Update tools modify existing records. The AI can update specific fields while leaving others unchanged.
Delete tools remove records from a datatable. These tools are marked with a destructive hint so the AI knows to confirm with the user before executing.
Buzzy Functions as Tools
If your app includes Buzzy Functions, these can also be exposed as MCP tools. This allows the AI to execute custom business logic you've defined, such as calculations, data transformations, or integrations with external services.
Widget Screens
Buzzy can generate widget screens that the AI displays to users. These are minimal UI components that show formatted data from your app. For example, a recipe app might have a widget that displays a recipe card with the title, image, ingredients, and instructions in a visually appealing format.
Connecting to ChatGPT
To connect your Buzzy app to ChatGPT, you need ChatGPT with Developer Mode enabled. In the Buzzy workspace, navigate to your app's MCP settings to find your MCP Server URL. Add this URL to ChatGPT's Developer Mode settings, and ChatGPT will discover the available tools from your Buzzy app.
When you first interact with your app through ChatGPT, you'll be prompted to authenticate via OAuth. This ensures that ChatGPT only accesses data you're authorized to see. After authentication, you can start asking ChatGPT to interact with your app using natural language.
Security and Authentication
Buzzy's MCP implementation uses OAuth 2.0 for authentication. Each user authenticates with their own Buzzy credentials, ensuring they only see data they have permission to access. The MCP server respects all of your app's existing security settings, including Viewers fields and team-based access controls.
Tool annotations provide additional security hints to AI clients. Destructive operations like delete are marked accordingly, prompting the AI to confirm with users before executing. Read-only operations are marked as safe, allowing the AI to execute them without additional confirmation.
Use Cases
Buzzy's MCP implementation enables a wide range of use cases. A meal planning app could let users ask ChatGPT to suggest recipes based on available ingredients, create shopping lists, and plan weekly menus. A project management app could allow users to create tasks, check project status, and get summaries of recent activity through conversation. A CRM could enable sales teams to query customer information, log interactions, and get insights without leaving their AI assistant.
The combination of Buzzy's no-code app building with MCP connectivity means anyone can create AI-powered applications without writing code.
Last updated