CI&T Data & AI Practice

Data & AI Strategy

CI&T Data Practice

A single Claude Code plugin that packages the full CI&T data & AI delivery chain. Install it once, get 17 specialized agents, 43 reusable skills, 28 MCP servers connecting to live systems, and 10 operational commands — all composable, all inside the IDE. Agents activate from conversation, invoke skills for workflow logic, and call MCP servers to reach live data. This portal is a deep-dive into the MCP server layer.

Our Mission

You don't start
from zero —
reuse and adapt.

Data projects fail not because engineers lack skill, but because each team rebuilds the same patterns from scratch. The CI&T Data Practice plugin puts a curated, battle-tested delivery chain inside every engineer's IDE — so the institutional knowledge of the practice is always one conversation away.

🧠

Agents over documentation 17 specialized personas that activate from conversation — no wiki to search, no template to adapt.

🔁

Skills over one-off scripts Reusable, agent-agnostic workflows with spec-review gates so every delivery follows the same standard.

🔌

Live data over context switching 28 MCP servers bring real system data into the conversation — no copy-paste, no stale screenshots.

⚡

Consistency at delivery speed From assessment to engineering to governance — one plugin, one standard, across every engagement.

What is the plugin? — Four building blocks

🧠

Agents

Who is acting

A persona with expertise, tone, and routing logic. Activates from conversation.

Data Architect Data Engineer BI Developer FinOps Data Strategist Data Scientist

⚙️

Skills

What to do

A reusable workflow with structured steps, spec-review gates, and artifact outputs.

data-discovery data-architecture data-engineering cost-analysis data-governance data-visualization

🔌

MCP Servers

What to access

Live connections to external systems — databases, cloud platforms, BI tools — from inside the IDE.

SQL Server Databricks Power BI Neo4j BigQuery Fabric

🛠️

Commands

How to operate

Slash commands for team workflows — install, ship, review PRs, set up MCP servers.

/init /ops:ship /ops:mcp-setup /ops:review-pr /welcome /ops:checkpoint

$ /cit-data-practice:data-discovery /cit-data-practice:data-visualization /cit-data-practice:cost-analysis /cit-data-practice:data-governance /cit-data-practice:data-assessment

/cit-data-practice:welcome /cit-data-practice:ops:ship /cit-data-practice:ops:review-pr /cit-data-practice:ops:mcp-setup

MCP is always optional. The plugin works end-to-end with local files (DDLs, exports, code). MCP is an enhancement for live systems — never a hard requirement. If no MCP is configured, the full migration pipeline still runs.

Agent Constellation — 17 personas, grouped by what they do

Orchestrator · Supervisor · Multilingual

oliver-concierge

The multilingual front door of the plugin. Oliver understands the user's goal, ranks matching assets by confidence, and routes to the right agent or skill with explicit user approval before executing anything.

routes all agents ↓

↓routes users to

Non-technicalshape the project — business, product, delivery

Pre-Engagement

sales-specialist

RFP responses, pitches & proposals

Strategy

data-strategist

Maturity, roadmaps, operating model

Product

data-product-manager

Specs, contracts, SLOs, adoption

Delivery

agile-delivery-lead

Delivery visibility: milestones, issues, cadence, RAID

Core data deliveryplans it, builds it, makes correct enforceable and governed

Architecture

data-architect

Migration pipeline owner — discovery to waves

Execution

data-engineer

Wave conversion + pipelines & quality tests

Quality

data-quality-engineer

6 DQ dimensions, quarantine, reconciliation

Governance

data-governance-architect

UC audit + design — grants, lineage, PII, glossary

Consumptionone or more, per project

Visualization

bi-developer

Power BI / Fabric semantic models, DAX — built on the certified measures

Conversational AI

ai-developer

Metric Views, Genie, natural-language querying

Decision App

app-developer

Databricks / Fabric Apps — a person decides, the app logs it

Data Science

data-scientist

Modeling, experiments, notebooks

MLOps

ml-engineer

Model serving, registry, drift monitoring

Advisory & auditengaged as needed — they assess and optimize, they do not build

FinOps

finops-architect

Databricks + Azure + GCP cost, waste detection

Cloud

cloud-architect

Multi-cloud infra, Well-Architected

Security

security-architect

Threat modeling, zero trust, compliance

Connects to your entire stack

28 MCP servers across 7 categories — credentials via vault / .env.

Lakehouse & Warehouse4

DatabricksSnowflakeMicrosoft FabricBigQuery

BI & Analytics2

Power BIMicroStrategyTableausoon

Source Databases5

PostgreSQLMySQLOracleSQL ServerNeo4j

Cloud & FinOps2

AzureGoogle CloudAWSsoon

DevOps & Delivery5

GitHubBitbucketAzure DevOpsJiraOpenText Octane

Workspace & SaaS9

DriveDocsSheetsSlidesGmailChatAnalyticsFormsNotion

AI & Agent Platforms1

Azure AI Foundry

Built on Claude — portable by design

The plugin runs in Claude Code today — but what makes it work isn't tied to one vendor. Agents are personas, skills are tool-agnostic workflows, and MCP is an open standard adopted across the industry. As other agentic environments mature, the same assets travel with them.

Agentic environments

Claude Code

Cursor

GitHub Copilot

OpenAI Codex

Google Antigravity

Enterprise AI platforms

Databricks Genie

Azure AI Foundry

Vertex AI

Amazon Bedrock

What it produces

.html

Reports & Portals

Discovery · Governance · Cost · BI

Self-contained HTML artifacts — heatmaps, radar charts, dependency graphs, and interactive data-science charts (Plotly/Altair/Bokeh). No server required.

data-discovery · data-governance · cost-analysis · data-science

.py

Python & PySpark

Pipelines · Notebooks · Models

Wave-by-wave PySpark conversion (tables, stored procs, SSIS, PowerCenter, Alteryx, copybooks) plus data-science pipelines — fitted models (.pkl), comparison CSVs, and synthetic datasets.

data-engineering · data-architecture · data-science

.sql

Migrations & Models

Stored Procs · dbt Models · DDL

Transpiled stored procedures, dbt model SQL, and DDL scaffolds for the target lakehouse schema.

data-engineering · data-architecture

.yml

Contracts & Config

Data Contracts · dbt Schema · DLT

Machine-readable data contracts, dbt schema.yml with tests, and DLT pipeline declarations.

data-engineering · data-architecture

.pbip

BI Package

Semantic Model + Reports

Deterministic PBIR 2.8.0 packages — semantic model, DAX measures, parity-mapped visuals ready to deploy.

data-visualization

.json

Dashboards & Specs

AI/BI Dashboards · Specs

Databricks AI/BI (Lakeview) dashboard definitions authored by the deterministic emitter, plus discovery scan results and reviewed spec artifacts.

data-visualization · data-discovery

.md

Docs & Assessments

Assessments · Architecture · Scan Docs

Assessment deliverables, architecture and wave plans, and one Markdown doc per scanned file — inputs, outputs, functions, and data elements.

data-assessment · data-architecture · data-scan

cypher

Knowledge Graph

Neo4j Nodes & Relationships

Source schema exported as a property graph — tables, columns, dependencies, and lineage edges queryable via Cypher.

data-discovery