For the complete documentation index, see llms.txt. This page is also available as Markdown.

Overview

Use this page when your app needs more than standard screens, fields, and actions. Buzzy can be extended and governed through code widgets, functions, constants, REST APIs, MCP, Tests, Security Review, toolkits, cookbooks, and widgets.

Choose the Right Extension Point

Need
Use

Custom UI, charts, maps, advanced filters, file managers, or embedded tools

Server-side logic, external APIs, webhooks, AI calls, or sensitive processing

External systems reading or updating Buzzy app data

AI assistants interacting with an existing app as tools

An AI agent building, adapting, or testing the app itself

Repeatable workflow checks and regression confidence

Permission, Private Data, and sensitive-data review before release

Shared UI layouts, app patterns, widgets, and reusable AI build guidance

Future in-app agent workflows

For a side-by-side comparison of code widgets, Async API, REST API, Node.js client, Buzzy Functions, and Custom MCP, see API and Extension Surfaces.

  1. Model the core app first: brief, flows, data, screens, and permissions.

  2. Use approved toolkits and cookbooks where they exist so the app follows shared layout, workflow, and governance patterns.

  3. Add code widgets only where standard Buzzy components are not enough.

  4. Move secrets and reusable configuration into constants.

  5. Put external calls and privileged logic in Buzzy Functions.

  6. Use REST API or Custom MCP only after the app's access rules are clear.

  7. Run Security Review for apps with sensitive data, role-based access, APIs, MCP exposure, or external integrations.

  8. Create Tests for the critical paths you need to trust before publishing or promoting the app.

Agent-Assisted Development

Agents such as Codex or Claude Code are useful for planning, writing code-widget internals, drafting Buzzy Functions, and running Playwright or browser-based checks. For app-building work, connect them through Buzzy Builder MCP so they work with Buzzy artifacts rather than guessing from screenshots or prompts alone.

For code widgets and functions, ask the agent to keep the runtime contract explicit:

  • which table, screen, row, or action invokes it

  • which fields or constants it depends on

  • which errors users should see

  • how it should be tested

Governed Extension Workflows

New extension work should be reviewed in the same context as the app definition:

  • Brief: confirm the app intent, roles, data, external functions, and security posture are still accurate.

  • Toolkits and cookbooks: use approved shared patterns before inventing app-specific structure.

  • Security Review: check Private Data, datatable permissions, field-level access, and external exposure.

  • Tests: capture repeatable checks for the workflows most likely to break or create risk.

Coming Soon

Buzzy Agents are coming soon. Until dedicated docs are available, use Buzzy Agents as the placeholder page and use Builder MCP for agent-assisted app building and Buzzy Functions for app-side automation.

Example Paths

Last updated