We recently announced the General Availability of KDB-X, the next-generation of kdb+ that unifies time-series, vector, and AI workloads in one high-performance environment.
Before the release, I hosted a live developer session with Michael Gilfix (Chief Product and Engineering Officer) and Andrew Wilson (Head of Engineering). Together, we looked under the hood of KDB-X GA and explored how this release expands what developers can build, connect, and deploy with KX technology.
Watch the full livestream below:
In this session we covered:
- The new KDB-X module system, which makes it easier to package, share, and reuse functionality across projects.
- Native support for open formats like Parquet and direct querying across object storage.
- Integrated Python, SQL, and q workflows for unified data and AI pipelines.
- The Model Context Protocol (MCP) server, which connects KDB-X to AI assistants and LLMs for natural-language querying.
- GPU acceleration for data-intensive workloads such as risk, research, and backtesting (still in development – but coming soon).
- Built-in dashboards for interactive, real-time visualization.
Whether you are building production systems, experimenting with AI pipelines, or just exploring the Community Edition, this conversation offers a closer look at what’s now possible in GA — and how the developer experience has evolved since the public preview.
Key takeaways from the conversation
1. A unified, developer-first platform
KDB-X brings together time-series, vector, and AI workloads inside one runtime so developers can work across data types and languages without leaving the same environment. It combines the speed of q with the accessibility of Python and SQL, making it easier to move from prototyping to production.
This GA release delivers a simplified installation process, first-class documentation, and a fully featured Community Edition that is free to use commercially. For developers, that means faster onboarding, fewer barriers to experimentation, and a single toolkit for building high-performance data and AI applications
2. The KDB-X module system
At the core of this release is the module framework, which provides a structured way to build and share functionality. Each module has its own namespace and export definitions, so developers can import code cleanly without name collisions or side effects.
This design allows code to scale from a simple library to a full ecosystem of components.
For example:
- Teams can version and share internal utilities without duplication.
- Partners can publish open-source libraries for specialized analytics or connectors.
- Developers can mix q, k, and C code in one environment.
Core modules released at GA include:
- Parquet – enables direct querying of Parquet files as native tables, removing the need for conversion or ETL.
- AI libraries – deliver hybrid vector and time-series similarity search for use in signal analysis, anomaly detection, and semantic retrieval.
- Object storage – allows developers to query data directly in S3, Blob, or GCS, combining cloud-scale data with in-memory performance.
- REST server and client (kURL) – make it easy to expose q functions as APIs or consume external services directly from within KDB-X.
Together, these modules turn KDB-X into a composable development platform where functionality can be extended or swapped with minimal friction.
3. MCP server connects KDB-X with AI
The Model Context Protocol (MCP) server integrates KDB-X with AI clients such as Claude, GPT, and Copilot. It allows large language models to query and interact with live or historical data stored in KDB-X using natural language, while maintaining full enterprise governance.
This bridges the gap between structured and unstructured data. Developers can combine time-series analytics with document or vector search in the same runtime, exposing that capability safely to AI agents. For data engineers and quants, that means users can ask complex questions in plain English and receive accurate, governed responses drawn from production data.
4. GPU acceleration for next-generation workloads
KDB-X GA introduces GPU-accelerated processing that can dramatically reduce compute time for both structured and vector workloads. Operations such as joins, sorts, and aggregations can now be executed on GPUs with minimal code change, using simple markers in scripts to define what should run on the GPU.
Early internal benchmarks show speedups of up to 25 times, allowing heavy workloads like backtesting, risk simulation, and large-scale model scoring to run in near real time. This capability helps teams compress batch windows, perform intraday analysis, and explore new classes of analytics that were previously impractical on CPUs alone.
5. Dashboards and real-time visualization
KX Dashboards are now fully integrated into KDB-X, allowing developers and analysts to visualize both streaming and historical data without needing a separate tool. Dashboards can subscribe directly to live feeds, update in real time through WebSockets, and connect seamlessly to q or SQL queries.
This integration transforms KDB-X from a data engine into a full insight platform. Developers can build and deploy visual interfaces that respond instantly to market data or system events, enabling real-time monitoring, research, and operational intelligence.
6. Open by design and built for growth
KDB-X embraces open standards such as Parquet, Arrow, and REST, making it easier to plug into existing infrastructure. Python, q, and SQL all operate over the same in-memory model, ensuring that data pipelines stay consistent and code remains reusable across teams.
The roadmap continues to focus on developer experience, with upcoming support for module package management (similar to pip for q), native testing and profiling tools, and even deeper fusion integration with external libraries. This open approach ensures that developers can extend KDB-X in any direction — across clouds, frameworks, and data formats — while keeping the same performance and simplicity.
Getting started
KDB-X GA represents the culmination of community feedback and continuous delivery since preview. It is not just a faster database. It is a unified, modular, and AI-ready platform that shortens the distance from idea to production.
You can try KDB-X today through the KX Developer Center. The Community Edition is free for commercial and offline use, and comes with tutorials, sample modules, and open-source integrations to help you get started quickly.
Start building: developer.kx.com/products/kdb-x/install
