Graphwise enhances its graph database to become the brains of AI agents

The Bulgarian graph database startup Graphwise today announced a major upgrade to its flagship GraphDB tool, adding new features aimed at boosting enterprise knowledge management and creating a more reliable foundation for artificial intelligence models.

Graphwise, officially known as Ontotext AD, said the improvements to its platform will make it even more suitable for AI applications, enabling them to access more knowledge from third-party sources and handle new use cases while also reducing infrastructure-related costs.

GraphDB is a specialized “graph database” that makes it possible to store contextual data alongside business records, such as information about which store a sale took place in. With traditional Structured Query Language databases, this isn’t always possible because their architecture doesn’t allow for these connections to be discovered easily. Graph databases, on the other hand, allow for connections to be stored alongside traditional records, facilitating more rapid analysis and queries.

According to Graphwise, the GraphDB platform is designed to address the gap between raw data and the need for actionable knowledge that can enhance the capabilities of large language models. So instead of treating content and data as separate concerns, it creates an intelligent fabric that connects them together.

To do this, GraphDB is based on a “knowledge graph” that provides a dynamic, interconnected view of enterprise content and how it relates to the database records. It works by encoding relationships, hierarchies and metadata in such a way that AI models can derive user’s intent, suggest related topics and support scenario-based recommendations. For instance, documents will be auto-tagged with both user-specified taxonomies and content-derived semantics.

The key element here is GraphDB’s so-called “semantic layer,” which is uniquely able to manage both structured data and unstructured content, while maintaining consistent semantic metadata that ties this information together.

New features in GraphDB 11

With the latest release, GraphDB 11, Graphwise says it’s making it easier for users to integrate the knowledge graph with multiple LLMs, streamlining knowledge access and adding support for the Model Context Protocol, paving the way for the creation of “agentic AI ecosystems” that can work autonomously.

GraphDB 11 adds support for a much wider range of LLMs, including models such as Meta Platforms Inc.’s Llama, Google LLC’s Gemini, DeepSeek Ltd.’s R1 and Alibaba Holdings Ltd.’s Qwen. There’s also an improved retrieval-augmented generation tool, known as GraphRAG, which facilitates access to enterprise knowledge bases to improve the accuracy of AI model responses.

By supporting the MCP protocol, Graphwise makes it possible for AI agents to tap into its knowledge graph, so they can be grounded in more relevant domain data. The result will be more reliable and accurate AI agents capable of providing context-aware insights, improving their decision-making abilities and boosting their efficiency across workflows, the company said.

In addition, the company has made great efforts to improve its “precision entity linking.” By this it means it has improved the way GraphDB maps terms and phrases inputted by users to the correct concepts or entities in its knowledge graph. This, it said, will help to eliminate ambiguity and improve how information is retrieved and applied to AI model’s outputs.

Besides boosting its capabilities, Graphwise is also trying to help users reduce their AI infrastructure costs, such as by adding support for the GraphQL language to simplify data access. The startup also claims higher availability with the new release, and performance optimizations such as advanced repository caching that boost its scalability and increase its responsiveness to make it more cost-effective.

Constellation Research Inc. analyst Michael Ni told SiliconANGLE that GraphDB 11 is a promising release that raises the bar in terms of what AI-ready enterprise data must deliver.

“By combining knowledge graphs with semantic reasoning and simplifying LLM integration and orchestration, Graphwise delivers a trust layer that grounds AI in how your business works,” the analyst said. “This isn’t just another graph release — it’s a rising blueprint and a delivered simplification for context-aware, decision-ready AI that we will see other enterprise platforms follow.”

Graphwise President Atanas Kiryakov said GraphDB 11 will help to solve challenges around the lack of “AI-ready data” and reduce instances of AI project abandonment. He said that around 60% of existing AI projects will ultimately fail and be canceled because the underlying models cannot access the data they need to deliver the anticipated results.

“GraphDB 11 directly addresses this by delivering the data infrastructure and governance that’s essential for cutting-edge AI,” he promised. “We empower customers to build intelligent, scalable applications by making their most complex unstructured data accessible and actionable.”

Image: SiliconANGLE/Dreamina