ZMDM + Claude Code Architecture | How They Work Together
🏗️ Technical Architecture

How ZMDM and Claude Code Work Together

A layered architecture where Claude Code acts as the always-on AI implementation team — generating configurations, writing scripts, and orchestrating setup — while ZMDM provides the production-grade Master Data platform underneath.

ZMDM — Master Data Platform
Claude Code — AI Implementation Agent
Business-Led, No IT Backlog

The Five-Layer Architecture

Claude Code (AI agent) operates as a Master Data Architect and Exprt that reads, writes, and orchestrates ZMDM — sitting above the platform layer and working alongside business users

Business Users
🧑‍💼
Data Stewards
Approve, reject, and enrich master data records via ZMDM forms
📦
Product / Supply Chain Teams
Submit new material, customer, and supplier requests through ZMDM workflows
📊
Data Owners & Governance Leads
Monitor quality scores, approve exceptions, manage policy via ZMDM dashboards
🔍
Zora AI Assistant
Natural language queries and guidance — Claude-powered, built into ZMDM
Instructs & Generates Config via API / CLI / MCP
Claude Code
🤖
Claude Code — AI Implementation Agent
Reads ZMDM schemas, generates data models, writes scripts, builds templates, tests integrations, and deploys configurations — all via command line and API. Acts as a tireless, expert MDM architect.
⌨️
CLI Commands
Direct ZMDM API calls to create objects, configure workflows, trigger migrations
🔗
MCP Server
ZMDM MCP integration lets Claude Code read schemas, list objects, and write config in real time
📝
Script & Template Output
Generates Python, JavaScript, and shell scripts — deployed to ZMDM or ERP integrations
Configures, Reads Schema, Pushes Data & Rules
ZMDM Platform
🏗️
Metadata Engine
Object definitions, field configs, data models — all Claude-readable and writable via API
⚙️
Workflow Engine
Configurable approval chains, escalations, SLA rules — set up by Claude Code, run by business
Quality Engine
Validation rules, completeness scoring, duplicate detection — rules authored by Claude Code
🔐
Governance & Audit
E-signatures, audit trails, role-based access, compliance controls — configured and verified by Claude Code
🌐
REST API / MCP Layer
Full programmatic access — the interface Claude Code uses to read, write, and test all ZMDM configuration
Reads / Writes Master Data Records
Master Data
🧱
Material Master
SKUs, BOMs, specs, classifications — governed and enriched by ZMDM + Claude Code
🏢
Customer Master
Account records, credit terms, ship-to locations — validated end-to-end
🚚
Supplier / Vendor Master
Approved vendor records, payment terms, compliance status — governed centrally
📍
Location / Plant Master
Sites, storage locations, organisational units — synchronised across systems
🔢
Reference & Classification Data
Units of measure, costing, tax codes, product hierarchies — single source of truth
Bidirectional Sync via Native Connectors
Source Systems
🏭
SAP S/4HANA & ECC
Native ZMDM connector — MM, SD, PP master data in sync
🔧
PLM / PDM Systems
Product specs, engineering BOMs — pulled in and governed through ZMDM
💼
CRM / Salesforce
Customer and account records — mastered in ZMDM, pushed to CRM
☁️
Cloud / Data Lakes
AWS, Azure, Snowflake — ZMDM as the master record, clean data flowing out
🔁
Other ERPs & Downstream Apps
Oracle, Microsoft Dynamics, WMS, MES — integration scripts authored by Claude Code
Key principle: Claude Code never becomes a dependency in production. It operates in implementation and configuration mode — all outputs are native ZMDM configuration, scripts, and data — so the business runs entirely on ZMDM once deployed.

Four Ways Claude Code Works with ZMDM

Claude Code interacts with ZMDM through four distinct patterns — each delivering a different kind of value across the implementation and operations lifecycle.

⚙️

Schema & Configuration Generation

Claude Code reads your existing data landscape and generates ZMDM-ready configuration — instantly.

1
Claude Code reads existing data sources — SAP exports, spreadsheets, legacy system extracts — via file upload or API
2
Claude Code generates a ZMDM object definition — field names, data types, mandatory flags, validation rules — as a JSON or YAML schema
3
Claude Code calls the ZMDM REST API or MCP server to push the object definition directly into the platform
4
ZMDM creates the configured data domain — forms, fields, and governance rules — ready for business users in minutes
🔌

ERP Integration & Data Pipeline Setup

Claude Code writes the custom field mappings and transformation logic that connect ZMDM to SAP and other ERPs.

1
Claude Code reads the SAP field catalog or ERP data dictionary — identifying all source fields and their business meaning
2
Claude Code generates a field mapping specification — ZMDM field → SAP MARA field — with transformation logic for unit conversions, code mappings, and defaults
3
Claude Code writes integration scripts (.py or .js) that handle incremental sync, error handling, and retry logic
4
ZMDM activates the connector using Claude Code's mapping config — real-time bidirectional sync begins immediately
🧹

Data Migration & Quality Cleansing

Claude Code profiles, cleanses, and migrates legacy master data into ZMDM — automating what normally takes weeks of manual effort.

1
Claude Code profiles the source dataset — identifying nulls, duplicates, format errors, and outliers across all master data fields
2
Claude Code generates cleansing scripts that standardise values, resolve duplicates, fill gaps using business rules, and flag exceptions
3
Claude Code loads cleansed records into ZMDM via the bulk import API, with a full exception log for records needing manual review
4
ZMDM Quality Engine scores every migrated record — Claude Code generates a remediation plan for any records below the quality threshold
📈

Ongoing Enhancement & Custom Logic

Post go-live, Claude Code handles any new requirement — new fields, new workflows, new rules — without IT tickets.

1
Business user describes a new requirement in plain language — e.g. "Add a Dangerous Goods flag to Material Master with approval from EHS team"
2
Claude Code reads the current ZMDM object schema via MCP, identifies the right place to add the new field and workflow step
3
Claude Code generates and pushes the updated configuration to ZMDM — new field, validation rule, workflow branch, and notification template
4
Change is live in ZMDM within minutes — business user tests in the UI, Claude Code writes a regression test to confirm nothing else broke

What Claude Code Does — and Doesn't Do

Claude Code is the implementation team, not the platform. It accelerates every phase of setup and enhancement, but ZMDM remains the system of record throughout.

What Claude Code Generates

  • ZMDM object and field schema definitions
  • Workflow configuration and approval chain logic
  • Validation and data quality rule sets
  • ERP field mapping and transformation scripts
  • Data migration and cleansing scripts
  • Integration test suites and regression checks
  • Exception and quality reports
  • User training documentation and process guides
🏗️

What ZMDM Runs in Production

  • Master data governance and approval workflows
  • Real-time data quality scoring and alerts
  • Native ERP and cloud system integrations
  • Business user forms, dashboards, and portals
  • Audit trails, e-signatures, and compliance controls
  • Zora AI assistant for business user queries
  • Automated enrichment and deduplication
  • Role-based access and data stewardship

What Moves Between the Systems

A precise view of every signal, artefact, and data payload exchanged between Claude Code, ZMDM, and the connected enterprise systems.

Signal / Artefact Direction Produced By Consumed By Purpose
Object Schema Definition → ZMDM Claude Code ZMDM Creates or updates master data domain (fields, types, constraints)
Workflow Configuration → ZMDM Claude Code ZMDM Defines approval steps, roles, SLA rules, escalation logic
Validation Rule Set → ZMDM Claude Code ZMDM Data quality gates applied on record submission and update
Current Schema / API Response ← ZMDM ZMDM Claude Code Lets Claude Code understand current config before making changes
Data Quality Report ← ZMDM ZMDM Claude Code Claude Code generates remediation scripts based on quality scoring output
Migration Script → ZMDM / ERP Claude Code Both Extracts, cleanses, and loads legacy master data into ZMDM
Field Mapping Spec → ZMDM Connector Claude Code ZMDM Tells ZMDM how to translate SAP / ERP fields to ZMDM data model
Master Data Records (synced) ↔ ERP ZMDM ERP / CRM / WMS Governed, enriched master data flowing into all downstream systems
Exception & Anomaly Report ← ZMDM ZMDM Claude Code Claude Code analyses and generates root cause explanations and fixes
Test & Regression Suite → ZMDM API Claude Code ZMDM Validates that new configuration changes haven't broken existing workflows
User Training & Process Docs → Business Users Claude Code Users & Stewards Auto-generated guides tailored to the specific ZMDM configuration deployed

How Claude Code Talks to ZMDM

Claude Code interacts with ZMDM through the REST API and MCP server — reading live schema, pushing configuration, and triggering data operations programmatically.

Claude Code → ZMDM REST API
# Claude Code: create a Material Master object # with validation rules via ZMDM REST API payload = { "domain": "material_master", "fields": [ { "name": "material_number", "type": "string", "required": True, "unique": True }, { "name": "base_unit_of_measure", "type": "reference", "ref_domain": "uom_master" }, { "name": "mrp_type", "type": "picklist", "values": ["PD", "VB", "ND"] } ], "workflow": "two_step_approval", "quality_threshold": 92 } response = zmdm_api.post( "/domains/configure", payload ) # → ZMDM domain created in < 1 second
Claude Code ↔ ZMDM via MCP Server
// Claude Code queries ZMDM schema live // via Model Context Protocol server const schema = await mcp.call({ server: "zmdm-mcp", tool: "get_domain_schema", params: { domain: "customer_master" } }); // Claude Code reads it, then generates // a new required field + approval step await mcp.call({ server: "zmdm-mcp", tool: "add_field", params: { domain: "customer_master", field: "credit_class", type: "picklist", values: ["A", "B", "C", "D"], approver: "finance_team" } }); // → Field live in ZMDM, no restart needed

Architecture Principles That Guide the Integration

Four principles ensure the ZMDM + Claude Code architecture is safe, scalable, and entirely business-owned in production.

Claude Code is Stateless in Production

Once configuration is generated and pushed to ZMDM, Claude Code plays no role in the live system. All outputs become native ZMDM configuration — workflows, schemas, rules — that run independently. There is no runtime AI dependency.

ZMDM is Always the System of Record

Master data records are created, governed, and stored entirely within ZMDM. Claude Code reads and writes ZMDM's API — it never holds data directly. All lineage, audit trails, and governance decisions are captured inside the platform.

Human Review Gates on High-Risk Changes

Claude Code can propose schema changes, workflow updates, and migration scripts — but business owners or data stewards review and approve before any change is promoted to production. The AI accelerates; humans decide.

Incremental, Testable Configuration

Every change Claude Code makes is versioned and tested via the ZMDM API before go-live. Claude Code generates regression test suites alongside every configuration change, ensuring new additions don't break existing processes.

Start Your Success Story

Join the growing list of manufacturers who have transformed their master data management with ZMDM — now supercharged with Claude Code.