Having previously served as Kurrent’s Technical Support Manager where he specialized in event-driven architectures and CQRS implementations, Lokhesh brings expertise hands-on experience to build complex event sourcing systems with cutting-edge AI integration.
This conversation is for those who love insights on AI-driven database solutions and learning about what organizations can do to ready themselves for the changes that are coming in the database automations.
M.R. Rangaswami: How is Kurrent’s MCP Server reimagining database interactions for both technical and non-technical users?
Lokhesh Ujhoodha: There’s a broader shift in how people interact with complex systems using MCP Servers. Traditionally, working with a database has meant writing structured queries, managing schemas and understanding low-level details like indexing or projection logic. That’s fine if you’re a backend engineer or database administrator, but it creates steep barriers for others who often have questions they can’t easily get answered without writing code.
With the introduction of MCP and emerging open-source servers, the way users interact with databases is starting to shift. Now users can interact with their data through natural language either directly with an AI assistant or through integrated tools. This means you can ask for a stream of events, write a new projection or debug logic through a conversational interface.
For technical users, this accelerates development cycles by enabling faster prototyping, allowing developers to test projection logic and debug issues through conversational commands. For less technical users, it lowers the barrier to engaging with data in meaningful ways. In both cases, it’s reimagining the interface layer between humans and systems by abstracting it into intelligent workflows that are easier to reason about and iterate on.
M.R.: Why is democratizing access to real-time data architectures so important as we look ahead to a future where agentic AI workflows become mainstream?
Lokhesh: The next generation of AI systems requires access to data that’s timely, contextual and complete. Unlike traditional AI that relies on static training sets or batched data, agentic workflows need to be fed with live signals and historical context in real time. Without that, their actions risk being either irrelevant or outright wrong.
This is why democratizing access to real-time architectures is such a key enabler. It’s not just about giving business users access to static dashboards or pre-built reports; it’s about giving both humans and AI systems the ability to interact with data in a feedback loop. For example, an AI agent that’s monitoring fraud can’t wait for a batch process to run overnight. It needs to see a suspicious transaction and contextualize it against previous behaviors instantly.
Most current data infrastructure, however, isn’t designed for that. It’s either streaming but not persistent, or it’s persistent but too slow or rigid for dynamic use cases. To support democratized AI participation across technical roles and business units, you need systems that treat data as a continuous stream of context-rich events, while still preserving fidelity and traceability.
This is where event-native solutions become particularly relevant. These systems combine immutable event storage with real-time streaming capabilities, enabling AI systems to access both live data and historical context instantly. Their event-native design can store billions of indexed streams with consistent ordering, providing complete data lineage while supporting real-time event processing. Multi-language support and simplified integration reduce the complexity barriers that often prevent broader adoption of real-time systems.
That’s what makes emerging real-time data platforms so significant. But because data streaming is not yet supported by MCP at this moment in time (although on the roadmap), the Kurrent MCP Server leverages the in-built real-time data processing projections and streaming capabilities of KurrentDB to also produce code that can get you started with live data subscriptions.
M.R.: What should organizations do to prepare for the shift from traditional database development practices to AI-driven automation, and what does this signal about the broader database market?
Lokhesh: Organizations preparing for AI-driven automation need to rethink how they store and access data. Traditional databases that only capture current state are essentially blind to the decision-making processes that AI agents require. When an AI agent needs to understand why something happened, how a system evolved or what patterns led to specific outcomes, point-in-time snapshots simply aren’t enough.
This is where storing complete state transitions becomes transformative for AI workflows. Unlike traditional approaches that discard the journey and keep only the destination, KurrentDB preserves every state change as a first-class citizen. This means AI agents have access to the full narrative of how systems evolved – not just what happened, but the sequence, timing and context of every transition.
This comprehensive transition history enables AI agents to perform temporal reasoning, understanding causality rather than just correlation. They can replay scenarios to test different decision paths, identify patterns across time that would be invisible in static data and learn from the complete audit trail of past actions. When an AI agent needs to make a decision, it’s working with the full context of how similar situations played out historically.
The future-proofing implications are significant. As AI capabilities advance toward more sophisticated reasoning, multi-agent coordination and causal understanding, having access to rich state transition data becomes exponentially more valuable. Today’s AI might use simple pattern matching, but tomorrow’s AI will leverage complete behavioral histories to make nuanced decisions, coordinate with other agents and provide explainable reasoning chains.
At a market level, this signals that the database layer is becoming the memory system for intelligent applications. The organizations that recognize this shift and adopt event-driven architectures with complete state transition preservation will have AI agents that can reason more effectively, learn more comprehensively and adapt more intelligently than those working with traditional state-only data models.
The competitive advantage isn’t just in having AI agents, but it’s in having AI agents with access to the complete story of how your systems behave and evolve over time.
M.R. Rangaswami is the Co-Founder of Sandhill.com