MCP Designer · Product direction

Describe the workflow. Build it with your AI assistant.

Open Codex, describe the business workflow, and ask it to prepare a draft in CXFabric. Then switch to Designer to inspect the nodes, mappings, policy, and approval path before anything is activated.

Natural-language brief Visual workflow Human review Source-backed guidance
Interactive product preview Chat first. Inspect the flow next.
chatgpt.com/codex
Codex
TaskBuild a CXFabric workflow
Connected to workspace
RF
You

Build a refund eligibility flow. Read the order, apply the 30-day policy, request approval above $250, then update the CRM.

Codex

I’ll use the CXFabric documentation and the components available in this workspace to prepare a reviewable draft.

Read CXFabric workflow documentation Checked available triggers, components, and connectors Created six-step draft with an approval branch

The draft is ready in CXFabric Designer. I left credentials, environment access, and activation unchanged.

Ask Codex to adjust the workflow...
Describe the work in chat Switch to the visual flow Review before activation

From prompt to visual draft

Keep the speed of conversation. Keep the clarity of a workflow.

The assistant helps translate a business request into the language of triggers, components, connectors, policy checks, branches, and system outcomes. The result is a workflow people can inspect together.

01

Describe the outcome

Explain the work, systems, rules, exceptions, and definition of done in plain language.

02

Ground the assistant

Use current CXFabric documentation, available components, and workspace context instead of model memory alone.

03

Draft the flow

Create the visual path, field mappings, branches, code steps, and implementation notes.

04

Review and activate

Let a person validate credentials, permissions, policy, test results, and production readiness.

AI assistance with boundaries

Let Codex prepare the work. Keep people accountable for it.

MCP Designer is intended to reduce repetitive design work, not remove engineering judgment. Every proposed change should remain inspectable in the Designer and pass through the same testing and operating controls as a manually built flow.

AI assistantPrepare and explain
  • Translate the requirement into workflow steps
  • Find relevant components and connectors
  • Draft mappings, logic, and code
  • Explain assumptions and source material
Human ownerApprove and operate
  • Choose credentials and environment access
  • Confirm policy and authority boundaries
  • Validate tests and exception behavior
  • Activate, monitor, and support the flow

Two MCP experiences

Design time and runtime solve different problems.

The existing MCP Trigger lets an MCP-capable client start a known CXFabric flow. MCP Designer is the product direction for helping an assistant create and revise the flow definition itself.

Product direction

MCP Designer

Works with the developer during design. It helps inspect context, propose changes, and prepare a reviewable visual workflow.

Conversation → workflow draft
Available today

MCP Trigger

Works at runtime. An approved MCP client connects to the generated endpoint and starts the flow associated with it.

MCP tool call → flow execution

Help shape MCP Designer

Bring a workflow you build today. Show us how you want to build it tomorrow.

We are exploring how MCP-capable assistants can make CXFabric workflow development faster without weakening review, security, or operational ownership.