This case study shows how the Model Context Protocol (MCP) enables artificial intelligence agents to automate complete ETL flows without the need to develop manual scripts. By exposing data pipelines as structured tools, MCP enables agents to extract, transform, and load data simply by following natural language instructions. This approach reduces integration complexity and facilitates the transition from code-based pipelines to automations orchestrated by AI agents.
Practical example: Keboola MCP Server in action. The Keboola MCP server converts Keboola pipelines into tools that can be invoked by AI. Agents can manage storage, execute SQL transformations, trigger jobs, and access metadata, all through natural language. An example of a natural language instruction could ask to create a daily transformation that segments customers who spent more than $100 last month, save the result to CSV, and update the dashboard. The server interprets the intent, generates the SQL transformation, schedules the execution, and monitors the process, delivering results and logs in structured JSON format to facilitate monitoring and error handling.
In practice, the agent can call operations such as create_transformation, run_job, or list_jobs and receive responses validated by Keboola. The remote connection supports SSE transport and OAuth authentication, enabling secure integrations from cloud clients or desktop agents.
Building a pipeline with natural language prompts. The agent receives a natural language instruction, translates the intent into operational steps, creates the SQL logic, schedules periodic executions, and manages the delivery of artifacts such as CSV files or dashboard updates. The MCP presents the results in a structured form, facilitating auditing and observability.
Cross-platform ETL: Confluent and Keboola. For hybrid flows, the Keboola and Confluent MCP servers work together. Agents can consume real-time Kafka topics from Confluent, clean and enrich data, and then route the information to Keboola for additional transformations and loading into a Delta Lake. Standard calls such as list_topics, consume_message, and run_transformation enable orchestrating processes across platforms through a consolidated MCP interface.
Agent orchestration. A single agent can coordinate real-time ingestion, scheduled transformations, and generation of deliverables such as CSV summaries or Power BI reports. The async-first architectures supported by Keboola and Confluent allow multiple workflows to run concurrently without blocking, ensuring high availability and performance in production environments.
Internal operation. Each tool exposed by the MCP servers includes metadata with name, description, input schema, and output format. The server validates inputs, executes the operation on the corresponding platform, and returns structured responses that facilitate error tracking. Logs are managed separately to keep JSON outputs clean while ensuring traceability and compliance.
Best practices and governance. Although MCP makes it easy to turn intentions into repeatable and auditable data operations, it is essential to establish controls. Limit write operations to reviewed tools, validate SQL logic through pre-execution checks, apply access control policies, and conduct log audits, especially in production environments. These measures balance automation, security, and speed.
About Q2BSTUDIO. Q2BSTUDIO is a custom software and application development company specialized in artificial intelligence-based solutions and cybersecurity. We offer custom software services, custom application development, AI agent integration, and AI consulting for businesses. We also provide AWS and Azure cloud services, infrastructure migration and modernization, and business intelligence services, including deployments and optimization with Power BI. Our experience enables us to design and implement ETL automations using AI agents, integrating MCP to orchestrate scalable and secure pipelines.
Featured Q2BSTUDIO services. Custom application development and custom software adapted to business processes, AI agent integration to automate data tasks, creation of automated ETL pipelines, cybersecurity consulting to protect flows and data, AWS and Azure cloud services for resilient deployments, and business intelligence services to turn data into decisions using Power BI. If your organization seeks to improve processes with artificial intelligence, AI agents, or modernize its data platform, Q2BSTUDIO supports you from definition to operation.
Conclusion. MCP transforms natural language instructions into reliable data operations, allowing teams to delegate routine data engineering tasks to AI agents. Implemented with appropriate governance controls, MCP brings automation, security, and agility to ETL processes. If you want to explore custom solutions or integrate AI agents into your data processes, Q2BSTUDIO offers practical experience in artificial intelligence, cybersecurity, AWS and Azure cloud services, and business intelligence services to take your project to production with guarantees.
References and recommended reading: sources on Keboola MCP, Confluent MCP, and AI agent use cases in ETL can be found in the technical blogs of Keboola, Confluent, and publications specialized in data engineering.
