Model Context Protocol (MCP)
MCP can mean two different Buzzy workflows. Choose the path that matches what you are trying to do.
In the Buzzy Workspace left navigation, MCP appears under Configure because it connects an app or build workflow to external AI clients. It is not a screen component by itself:
use Builder MCP when an AI coding agent should build, inspect, or edit the app definition
use Custom MCP when an existing app should expose governed app-specific data and actions to MCP clients such as ChatGPT or Claude
Both workflows still depend on the same app building blocks: brief, flows, data model, blueprint, screens, theme, functions, widgets, people, settings, and permissions.
Starting and Iterating with MCP
Builder MCP
Connect an AI coding agent to Buzzy, create or select an app, then work through staged artifacts such as brief, flows, data model, theme, blueprint, and screens.
Ask the agent to inspect the current app definition, make a bounded change, validate affected artifacts, update screens/functions/widgets, run checks, and produce reviewable changes.
Custom MCP
Enable MCP on an app that already exists and has a clear data model, permissions, functions, and security posture.
Review which datatables, CRUD tools, functions, and widgets are exposed; test with allowed and denied users; update the app definition as the assistant-facing workflow evolves.
Buzzy Builder MCP
Use Buzzy Builder MCP when an AI agent such as Codex or Claude is helping you build, template, inspect, or edit a Buzzy app.
This is the builder workflow: your agent connects to Buzzy's MCP server, follows the Buzzy bootstrap instructions, works through the app-building stages, and helps you review or refine the generated app.
Start here if you want to:
create a new app with an AI coding agent
start from an existing Buzzy template
edit an existing Buzzy app through an agent-assisted workflow
set up and bootstrap an MCP-connected builder repo
use review gates, screenshots, Playwright checks, and staged changes while building
New here? The hands-on Tutorial: Build a carer app from a template is the quickest way to see the whole Builder MCP workflow end to end.
Buzzy Custom MCP
Use Buzzy Custom MCP when you have a Buzzy app and want that app to be available to AI assistants such as ChatGPT or Claude.
This is the app-exposure workflow: Buzzy generates MCP tools from your app's datatables, functions, permissions, and screens so authenticated users can query data, create records, perform actions, and view widgets through an AI assistant.
Start here if you want to:
enable MCP for an app you built in Buzzy
let users interact with app data through ChatGPT, Claude, or another MCP-aware client
understand generated MCP tools, OAuth, widgets, and app security rules
test that REST API, runtime, and MCP access return the same allowed data for each user
Accessibility and verification
Builder MCP exposes structured tools that can inventory applicable native-widget checks, run a persisted screen-structure preflight, execute browser accessibility audits, retrieve repeated evidence and evaluate an exact-scope release gate. Custom MCP can provide a structured, non-visual path into an app, while the accessibility of the MCP client's own interface remains the responsibility of that client.
See Accessibility at Buzzy for the current support and test scope, the complete Builder MCP accessibility tool list, the recommended agent workflow and practical manual-testing guidance.
Shared MCP Concepts
The Model Context Protocol is an open standard that lets AI assistants interact with external applications and data sources in a structured way. In both Buzzy workflows, MCP provides the bridge between an AI client and Buzzy, while Buzzy handles app structure, data, permissions, and runtime behavior.
When you need the canonical definition of Buzzy row data, metadata fields, parent-child relationships, or linked-table values, refer to the REST API docs:
Last updated