New tool in Elasticsearch helps developers quickly develop and manage custom AI agents through natural language interaction.
Elastic introduces Agent Builder, a new functionality within Elasticsearch that enables developers to build custom AI agents based on their own enterprise data. The tool is now available as a preview in Elastic Cloud and focuses on simplifying the development and management of AI agents through natural language.
Focus on Context Engineering
Agent Builder addresses the growing demand for AI agents capable of handling complex and data-driven tasks. In many organizations, relevant context is spread across various unstructured data sources such as documents, emails, and customer data. Context engineering is essential to ensure accurate answers.
Within Elasticsearch, Agent Builder offers a range of features to manage this context more efficiently. Developers can, among other things, ask questions to their data in natural language, determine which data sources should be consulted, and configure how the agent interacts with that data. For this, the tool uses the Model Context Protocol and offers integration options for external applications via A2A connections.
Customization and Control
Agent Builder also includes capabilities to fully define AI agents. Developers can set custom system instructions, configure access rights, and build tools that give the agent new capabilities. All of this happens within the existing Elasticsearch environment, centralizing control over security, governance, and relevance.
Furthermore, Agent Builder supports advanced search techniques such as semantic and hybrid queries and utilizes the query language in Elasticsearch. This allows developers to precisely determine which data is offered to the underlying language model.
read also
Small and Large Language Models: are SLMs Small LLMs?
The introduction of Agent Builder is part of Elastic’s strategy to further integrate AI functionality within its search platform. Version 9.2 of Elasticsearch will soon make the functionality more widely available beyond the serverless preview.