Best MCP Servers for Data Engineering in 2026

AI + Data Engineering with MCP

Data engineers spend hours writing SQL, debugging pipelines, and exploring datasets. MCP servers give your AI assistant direct access to databases, notebooks, and analytics tools.

Here are 6 verified MCP servers from the Spark catalog for data work.

1. Jupyter MCP Server

Connect AI to Jupyter Notebooks - execute code, manipulate cells, and view outputs including charts and images in real time.

Use case: "Run the data quality checks in cell 5 and show me the output. If there are failures, add a new cell with fixes."

View on Spark →

Pros: Real-time cell execution · Multimodal output · Any Jupyter deployment (local, hub, cloud)
Cons: Requires running Jupyter server

2. ArangoDB Graph MCP Server

Advanced graph database operations - manage collections, run AQL queries, handle graph traversals, with backup/restore and analytics.

Use case: "Show me the shortest path between these two entities in the knowledge graph and list all intermediate connections."

View on Spark →

Pros: Async Python architecture · Graph traversal · Backup/restore · Content conversion
Cons: ArangoDB-specific

3. Fabric Real-Time Intelligence MCP Server

Microsoft Fabric's real-time analytics through AI. Run KQL queries on Eventhouse, manage Eventstreams, and configure Activator rules.

Use case: "Query the last 24 hours of user events, aggregate by event type, and show me a time series."

View on Spark →

Pros: KQL query execution · Real-time streaming · Full RTI coverage
Cons: Microsoft Fabric subscription required

4. AutoML MCP Server

Automated machine learning - data analysis, preprocessing, model selection, and hyperparameter tuning through MCP tools.

Use case: "Take this dataset, run feature analysis, select the best model for churn prediction, and show me the metrics."

View on Spark →

Pros: End-to-end ML pipeline · Automatic feature engineering · Model comparison
Cons: May need significant compute for large datasets

5. arXiv API MCP Server

Search and retrieve scientific papers through natural language. Find papers by topic, get metadata, download PDFs.

Use case: "Find the latest papers on graph neural networks for time series forecasting and summarize the top 3."

View on Spark →

Pros: Natural language search · Full metadata · PDF download
Cons: Read-only (can't submit papers)

6. Source to Knowledge Base MCP Server

Transform code repos and Notion content into searchable knowledge bases with AI-powered semantic search.

Use case: "Index our data pipeline repo and make it searchable. Then find all functions that handle null values."

View on Spark →

Pros: Semantic code search · OpenAI embeddings · Notion integration
Cons: Requires initial indexing step

Data Engineering Workflows

Schema exploration: "Connect to the graph DB, show entity types and their relationships"
Pipeline debugging: "Run the notebook cell by cell and find where the data shape breaks"
ML experimentation: "Try 3 different models on this dataset and compare accuracy"
Research: "Find recent papers on our approach and check if there's a better method"

Explore More

Browse all 875+ MCP connectors on Spark - databases, analytics, cloud storage, and more.