Today, we are happy to share the latest updates for Starburst Galaxy. This release advances our commitment to building a simple, easy-to-use data foundation based on Apache Iceberg + Trino on the fully managed Galaxy platform, accelerates the performance of our managed infrastructure, and delivers the next generation of agentic AI capabilities.
We are removing the operational tax of manual data management, so your team can move directly from raw data to insights in a single, simple package that skips both the ETL at one extreme and the BI layer on the other.
Starburst Galaxy moves beyond dashboards into the AI Era
Starburst Galaxy is evolving from a query engine into a platform for agentic AI. This enables users to interact with data through conversation rather than static reports. As the industry moves from static reporting to AI-driven decision-making, the limitations of traditional dashboards become impossible to ignore. Dashboards were designed for a previous era of fixed questions and historical summaries. We are now entering a new era characterized by continuous and conversational interaction with data.
The Agentic Interface Built on Galaxy
This release strengthens the layers that power AIDA, Starburst’s upcoming AI Data Assistant, designed to replace static dashboards with governed, conversational decisioning across all your data.
What can AIDA help you do?
AIDA introduces a new interaction model for enterprise analytics. Instead of rebuilding dashboards for every new question, users engage in conversational flows where one answer naturally leads to the next, deeper question. Analysis evolves dynamically through dialogue, moving at the speed of the business rather than the speed of a ticket queue.
This interaction model is only possible because AIDA is built on Galaxy’s shared organizational context. Rather than operating on raw tables, AIDA builds its reasoning capabilities on the back of curated, governed data products. As a result, definitions, ownership, lineage, and policies travel with every query and every result, ensuring that metrics mean the same thing across teams. At the same time, data governance is enforced by default, aligning answers to business logic from the outset.
AIDA combines Agnentic Control Plane and Starburst MCP Server
Under the hood, AIDA is powered by the Starburst AI Agent, a reasoning agent that goes beyond simple text-to-SQL. It interprets user intent, refines queries iteratively, and validates execution paths against governed data assets to deliver accurate, business-aligned insights.
Supporting this interface is Galaxy’s Agentic Control Plane, anchored by the Starburst MCP Server. The MCP Server provides a secure, standards-based gateway that allows both internal and external AI agents to interact with Galaxy as first-class users.
Through the MCP Server, agents can:
- Discover available catalogs, schemas, and tables.
- Sample data to understand structure and context.
- Execute analytical queries through a Smart Routing Endpoint that automatically selects. the appropriate execution path.
An Enterprise Intelligence Platform built for the future
Rather than generating blind SQL queries run on isolated systems, agents explore and reason over the full data estate using the same permissions and governance controls as human users. All agent access is authenticated, authorized, and auditable, ensuring enterprise-grade governance extends to your entire AI ecosystem.
Together, the Starburst AI Agent and MCP Server lay the foundation for AIDA, enabling an agentic interface built on top of Galaxy’s Icehouse analytics engine, federated access model, and governed Data Product layer.
Supporting AI with superior data ingestion and performance benchmarks
We have spent the last year dismantling the barriers to data ingestion and data maintenance to ensure your data is ready for action in near real time. Our vision is based on a platform where Apache Iceberg becomes the data foundation powering AI workloads.
To this end, our approach creates a platform for Iceberg ingestion and maintenance that removes barriers when using Iceberg for AI workloads.
Starburst supports AI workloads with high-velocity streaming and file ingestion
Starburst Galaxy already includes streaming ingest capabilities built into the platform, allowing users to ingest up to 100GB per second into managed Iceberg tables with built-in exactly once delivery and sub-minute latency.
Over the past year, we have also automated file ingestion from Amazon S3. This feature automatically discovers and schematizes massive volumes of data so you can load it directly into Iceberg without writing custom code.
Together, these features support the data foundation needed for AI. In this, Starburst as a platform helps enable workloads that support the agentic AI transition directly.
Independent benchmarking results prove that Starburst Galaxy outperforms competitors
Recent validation from Concurrency Labs confirmed the strength of our architecture. In head-to-head tests against AWS Data Firehose and Confluent Tableflow, Starburst Galaxy achieved approximately 7x the record ingestion rate. The cost efficiencies were equally significant. Benchmarks showed that Starburst ingestion costs were 87% lower than AWS Data Firehose and 81% lower than Confluent Tableflow at the tested throughput.
Expanding Format Support and Schema Registry
We are excited to announce the Public Preview of Avro support for streaming ingest, along with a new Confluent Schema Registry integration. These tools allow you to manage and evolve schemas while Starburst handles the heavy lifting of deserialization. On the file ingestion front, we are expanding beyond JSON to include Public Preview support for CSV and other common text delimiter types.
Transition from BI to AI today
These updates represent more than just a faster platform. They mark the end of the era of static dashboards and the beginning of agentic analytics.
By eliminating the manual burdens of infrastructure and Icehouse maintenance, Starburst Galaxy allows you to move beyond the constraints of traditional BI. You can now spend your time building the future of AI-driven decision-making instead of managing legacy tools.
