Accelerators

Accelerator · Fuuz package

Fuuz Agentic Workflows

Working examples of MCP tools built as Fuuz data flows — seven live tools, the chat interface that drives them, and the models, screens and configuration behind it. Sanitised from a live ISA-88 MES tenant.

Get the package All accelerators

Working examples of MCP tools built as Fuuz data flows, plus the chat interface that drives them — lifted from a live ISA-88 MES tenant and sanitised for publication.

A Fuuz flow becomes a Model Context Protocol tool by carrying an mcpToolConfiguration with enabled: true. Nothing else is required: no separate server, no adapter. Seven of the ten flows here are live tools; the other three are the conversation UI.

What's here
7 MCP tool flows DataFlow Details · DataModel Details · DataModel To Graph · Dataflow Diagram Flow · Dynamic Query · Screen Details Flow · i3x_write_value_aw
3 conversation flows Agentic Chat Web Flow and its page-load companion, plus AI Interface Dialogue Web Flow
4 data models Conversation, ConversationMessage, ConversationStatus, MagicSpreadsheet
3 screens Fuuz Agentic Interface, Agentic Response Dialogue, Fuuz AI Package Info
1 configuration System Prompts — the prompt guide and examples the flows draw on
1 connection template claude-api-endpoint.template.json

The MCP tools, and what they expose

These turn a Fuuz tenant into something an agent can inspect and query:

  • Dynamic Query — run a query against any model in the tenant
  • DataModel Details / DataFlow Details / Screen Details Flow — read the definition of a model, a flow or a screen
  • DataModel To Graph — the model graph, relations included
  • Dataflow Diagram Flow — a flow rendered as a diagram
  • i3x_write_value_aw — write a value through the i3X interface

An agent with these can answer questions about a tenant's structure without anyone exporting anything.

What has been removed

  • Credentials. No API key ships. connections/claude-api-endpoint.template.json is a template with a placeholder — create the connection in your own tenant. Keep the id claudeApiEndpoint if you want the flows to work unmodified.
  • User identities. 1,010 createdByUserId / updatedByUserId values were nulled.
  • Tenant identifiers. None remain.

Everything else — the full flow definitions with their node graphs, the model field definitions, the screen layouts — is intact and importable.

Using them

  1. Create the Claude API connection from the template, with your own key.
  2. Import the models, then the flows, then the screens — in that order; each layer references the one before it.
  3. Add the System Prompts application configuration.
  4. In each flow's mcpToolConfiguration, confirm enabled and give the tool a name and description your agent will see.

These are examples, not an accelerator

They are published to show how the pattern works, not as a supported package. They were built for one tenant's ISA-88 MES app and carry that shape. Read them, take the pattern, and expect to adapt the specifics.

Before you install

A Jump Start, Not A Finished Answer

An accelerator is a pre-built, configurable starting point that addresses the bulk of the common challenges in its domain. It is not intended to solve every possible variation off the shelf — requirements that fall outside it are expected and normal for the model, not a shortcoming of this one.

Standard

Delivered and working as-is. No configuration needed.

Configurable

A base capability you adapt with the platform's own no-code and low-code tools — screen builder, transforms, workflow rules, flow parameters — without professional services.

Custom development

Feasible on the platform but outside this accelerator's current scope. A scoped change through Fuuz or a certified partner.

Once installed, it is yours

Installation forks the flows, models and screens into your environment as your own copies. Later versions published against this accelerator do not propagate to you automatically — picking up a change means deliberately re-applying it.

How accelerators work