Inside the platform: capabilities, architecture and security.

4MDG is a 100% cloud-based SaaS MDM platform, AI-native: an assistant in every module, an MCP server that puts your data governance inside Claude, Cursor or any compatible host, low-code workflow, REST API with OAuth 2.0 and search resolved within the database itself. This page goes into the detail: each block opens in a drill-down from the business summary to the technical specification.

OCI · Cloud SaaSMCP · AI in your hostAPI REST + OAuth 2.0ISO 27001 and LGPDADAM · AI in every module

Model Context Protocol

Your data governance inside your AI.

The platform exposes a native Model Context Protocol (MCP) server. Claude, Cursor or any compatible host connects straight to your tenant and, with the permissions of whoever authorized the connection, you ask, register and configure in plain language inside your own AI assistant. For example: "how many suppliers have had a pending record for more than 5 days?".

Native MCP server

The platform exposes a Model Context Protocol server (HTTP streamable). Claude (desktop, web and Claude Code), Cursor or any MCP client connects straight to your tenant.

With that person's permissions, and nothing more

The connection is authorized by the user through OAuth. The AI sees and does exactly what that person can see and do on the platform, respecting roles, departments, stages and row-level access rules.

All six modules, not just queries

The same tools as the internal assistant are available in the host: query and aggregate records, open and fill in registrations, design models, flows and rules, test rules against real samples, manage members and permissions, configure deadlines, compare and promote environments, administer organizations.

Irreversible actions require human confirmation

Publishing a record, promoting an environment, removing access: over MCP the AI prepares the action and returns a link. The person confirms it, inside the platform.

Traceable

Everything that arrives through MCP is logged as done by the person via that AI host, and value suggestions carry the MCP host as their origin.

Multi-organization and multi-environment

Those working across several organizations or environments choose where each call applies, without switching logins.

160 clients trust us

NestléCarrefourGeneral MotorsHeinekenWhirlpoolHershey'sDasaSoftysJactoPanasonicIntelbrasiFoodNestléCarrefourGeneral MotorsHeinekenWhirlpoolHershey'sDasaSoftysJactoPanasonicIntelbrasiFood

Model

Multi-tenant SaaS cloud, with a subdomain per client

Cloud

Oracle Cloud Infrastructure (OCI), with Cloudflare global edge

Integration

REST API documented in OpenAPI, OAuth 2.0 and MCP server

Compliance

ISO 27001, LGPD and full history per record

Capabilities

The business designs the process.IT is not the bottleneck.

Every registration in 4MDG is a configured workflow, not written code. Open each block to see how this is implemented.

Fluxo de um workflow 4MDG: requisição, validação automática, enriquecimento por conector, aprovação, publicação do golden record e envio ao ERP, com histórico completo
Requisição, validação automática, enriquecimento por conector, aprovação, publicação do golden record e envio ao ERP, com histórico completo.

The platform on screen · 1 of 5

Registration in steps, with Personal Data and Address filled in side by side

01

Registration in steps

Named steps, filled side by side, with the time remaining for each and sections marked as Master Data or Process.

Six roles: Curator, Architect, Guardian, Inspector, Maintainer and Administrator

02

Six roles, six slices

Curator, Architect, Guardian, Inspector, Maintainer and Administrator: each role sees only what it needs to do its part, each with its own ADAM.

Tests, Localities and Enrichment domains, each with its own models, records and pending items in view

03

Domains and models

Each business area organized into domains and models, with records and pending items in view, and the environment in use always shown at the top when it is not production.

Dashboard of indicators and a filterable, paginated table of records

04

Report

Configurable charts per model and a record table with search and filters, without leaving the platform.

Conversation with ADAM answering questions about the database and proposing a draft change

05

ADAM answers and proposes

Ask your own database in plain language ("how many people are married?") and request a change: ADAM prepares the draft, and anything irreversible only happens when you confirm.

ADAM · AI in every module

AI in every module, with the decision always human.

ADAM is the platform's AI assistant: one per module, always with the permissions of whoever asks. It proposes, you decide, and everything it does is recorded as done by you.

One assistant per module

Curator, Architect, Guardian, Inspector, Maintainer and Administrator each have their own ADAM, with the tools for that job.

Query your data in plain language

Queries and aggregations over your own records, respecting the permissions of whoever asks.

Assisted registration

ADAM opens the draft, fills in fields and suggests values. Each suggestion states whether it was observed in a source, derived or unknown, with the evidence, and only becomes a value when a person accepts it.

Configuration by conversation

The Architect's ADAM designs models, flows and rules and tests those rules against real records; the Guardian's manages access; the Maintainer's compares and promotes environments.

System-driven selection

Via API, the AI chooses among candidates the record matching a description, used to match items to the standardized materials catalog.

Explainability, audit trail and spending cap

Each suggestion records its origin, evidence and who produced it. Everything the AI does stays in the history as done by the person, and irreversible actions require explicit confirmation. Each organization chooses its language model from a curated catalog and sets a monthly spending cap.

Ciclo do ADAM: consulta em português, cadastro assistido e configuração por conversa, sempre com a sugestão explicada e a confirmação da pessoa
Pergunta, sugestão com evidência e confirmação: onde a IA propõe e onde a pessoa decide.

Documentation and comparison

Docs and comparisongenerated from your live setup.

Every workflow and every model has its own documentation, always up to date because it comes from the living configuration of your environment, not from a separate manual.

Living documentation

Every workflow and every model has a documentation page generated from the current configuration, ready to print or save as PDF.

Comparison between environments

What was created, changed or is in conflict between a test environment and production, before promoting.

Promotion with segregation of duties

Whoever created the changes does not promote them alone.

ADAM explains the workflow

Ask the assistant how a workflow works, which fields a step depends on or what a rule does.

Architecture

A platform that handles the load of both peak and ordinary days.

Separate layers and asynchronous processing for heavy workloads, with per-organization isolation enforced at the database level. Open each block for the components.

Stack

OCI Kubernetes Docker Cloudflare PostgreSQL + busca BM25 Filas no PostgreSQL Isolation by organization CI pipeline REST / OpenAPI OAuth 2.0 MCP

Integration

Integration ready from the first workflow.

An OpenAPI-documented REST API, an MCP server and connectors we maintain connect the platform to your ERP from the first published workflow.

Public REST API

Documented in OpenAPI, with an interactive reference in five languages; GET, POST, PUT and DELETE. OAuth 2.0 authentication with dedicated service accounts, one per environment, revocable at any time.

Intake and query

An endpoint to open records in any workflow and an endpoint to read records from any model.

Connectors

Every call the platform makes to a connector is cryptographically signed (Ed25519); open SDK and template to build your own connector. The 4MDG connector brings enrichment automations based on public registry databases (Receita Federal, Simples Nacional, Sintegra).

Model 1

ERP response

Each submission records status, attempts, the ERP identifier and the return message, all visible in the record itself.

Model 2

4MDG → ERP

The platform sends data to the ERP through a connector. The flow originates in 4MDG and is directed to the ERP and legacy systems.

Systems and protocols

REST / JSON OAuth 2.0 MCP SSO SAML 2.0 / OpenID Connect SCIM 2.0 Microsoft Entra ID Okta Carga em lote por planilha Tempo real ERPs como SAP, TOTVS e Oracle via conector
4MDG como golden record no centro, com consulta à Receita Federal e ao Sintegra via conector à esquerda, e integração por API REST e servidor MCP com o ERP à direita
4MDG no centro, ERP à direita, fontes de consulta via conector à esquerda, com os sentidos de cada integração.

Security and compliance

Master data is sensitive data.We treat it that way.

Control at the edge, control in the session, control in the record. And documented proof of everything that happened.

ISO 27001

Certified information security management system, with third-party audited controls and periodic review.

LGPD and GDPR

Compliance with the LGPD and GDPR in the handling of personal data.

Encryption in transit

All access over HTTPS, with TLS 1.2 or higher at the Cloudflare edge, and a digital certificate issued by a valid certificate authority.

Authentication and SSO

Corporate SSO per organization, via SAML 2.0 or OpenID Connect, with identity provider groups becoming departments. Also available: login with Google, Microsoft, passkeys, and a mandatory second factor for password login, with automatic provisioning via SCIM 2.0.

Granular access control

Roles per module; permissions by department, flow, and stage (view, edit, approve, advance, publish); section-level visibility; row-level rules.

Complete history per record

Every change records who made it, when, through which channel (screen, API, assistant, MCP, bulk load), and the values changed, with comparable versions in the record itself.

AI under control

AI acts with your permissions, irreversible actions require confirmation, and everything it does is kept in the history.

Continuity

Security and recovery procedures supported by managed cloud services.

Data and analytics

Your data, in your BI, with controlled access.

The same base that powers the platform also feeds your analytics, with no separate index to keep in sync. Authorized users build searches, dashboards and reports on their own records, within the same permissions and the same audit trail as the master data.

Configurable charts. Per model (donut, bar, column, line; count, sum, average, minimum and maximum), clickable to filter the table.

Ask and get answers instantly. Ask ADAM or your own AI via MCP and get the aggregation right away.

Into your corporate BI. Records flow out through the REST API into your corporate BI.

Quatro cadastros do mesmo cliente, marcados como duplicado, incompleto e incorreto, passando por padronização, validação por regras e identidade por campos-chave até virar um golden record com nome, e-mail, telefone, endereço, CPF e status

Contact

Talk to our technical team about your master data architecture.

We review integrations, security requirements and your environment's architecture with you. No commitment.

+55 11 4113-2510 atendimento@4mdg.com.br

Av. Queiroz Filho, 1700, Tower E, Suites 715 to 718
Vila Leopoldina, São Paulo, SP, 05319-000

Your data is processed in accordance with the LGPD.

IT questions

The questions your technical team will ask first.

Schedule a technical call