Agent Builder now GA: Ship context-driven agents in minutes

Agent Builder is available now as a tech preview. Get started with an Elastic Cloud Trial, and check out the documentation for Agent Builder here.

We are excited to announce the general availability of Agent Builder in Elastic Cloud Serverless and in the upcoming 9.3 release. Agent Builder brings the power of Elasticsearch as a context engineering platform to quickly develop contextual, data-focused AI agents.

Agents are gaining traction driven by their potential to deliver efficiency gains and better customer experiences. But in practice, providing agents with the right context is difficult, especially when operating over messy, unstructured enterprise data. Developers must manage tools, prompts, state, reasoning logic, models, and crucially retrieve relevant context from business sources to deliver accurate results and actions. Elastic Agent Builder delivers these core components to develop secure, reliable, context-driven agents.

Agent Builder core capabilities

Agent Builder leverages Elastic’s long-term investments in search relevance and retrieval-augmented generation, and work to make Elasticsearch the best vector database to simplify the development of contextual, data-focused AI agents.

Agent Builder allows you to:

• Immediately start with a built-in conversational agent that can answer questions, perform analytics and drive investigations over any data in Elasticsearch.
• Quickly go from complex unstructured data to a custom agent with configuration-based development experience.
• Leverage best-in-class, hybrid search relevance through built-in ES|QL or custom tools to improve context quality and agent reliability.
• Execute complex workflows (preview) as reusable tools to enrich data, update records, send messages, and more for rules-based automation.
• Connect to data sources outside of Elasticsearch using workflows and MCP to correlate and combine context for agents.
• Integrate with any agentic or application framework using built-in and custom tools exposed over MCP, and the ability to connect to external MCP (preview), support for A2A, and full API support.
• Extend the capabilities of Agent Builder with integration to third-party solutions like LlamaIndex for complex document processing or Arcade.dev for secure, structured tool access.

To further extend Agent Builder functionality, we are introducing Elastic Workflows, our new rules-based automation capabilities, now in technical preview. For organizational tasks, agents at times need certainty and reliability of rules-based actions, which are often necessary to implement specific business logic. Elastic Workflows provides agents with a simple, declarative way to orchestrate internal and external systems to take actions, gather and transform data and context. Workflows are fully composable, event-driven and flexible, and can be exposed as tools to an agent via MCP.

Go from data to agent in minutes

Developing agents can take weeks of upfront work to consolidate separate data stores, build manual pipelines, tune queries, and manage complex orchestration. Agent Builder reduces the time to develop agents by removing the need for separate data stores, vector databases, RAG pipelines, search layers, query translators, and tool orchestrators, allowing you to focus on agent logic and application delivery.

Agent Builder natively integrates Elasticsearch platform primitives to make agent development fast.

• Start with a built-in conversational agent that can immediately chat with and reason with your indexed data.
• Integrate agents into applications, dashboards, or CI/CD systems with interactive access via Kibana, APIs, or MCP and A2A.
• Build with default tooling to understand your data structure, select the appropriate index, generate optimized hybrid, semantic, and structured queries, and create configurable visualizations using ES|QL based on natural language prompts.

To go deeper, try a complete hands-on walkthrough.

Build on Elasticsearch, a complete data platform for context engineering

For AI agents, context quality is essential to provide effective reasoning and reduce the risks of hallucination. For many enterprise AI agents, the business data required to perform a task is the most crucial piece of context. As a massively scalable data store, vector database, and leader in relevance, Elasticsearch already offers many strong context-engineering primitives. Context engineering goes beyond simple retrieval-augmented generation by allowing you to tailor and scale how data is fetched, ranked, filtered, and presented to agents, helping reduce noise and ambiguity.

Elasticsearch delivers a context engine that combines lexical search, vector search, and structured filtering for retrieval that materially improves LLM performance by ensuring the model operates on relevant and precise context. This capability is supported by agentic retrieval, along with built-in tools and search logic that automatically select the right indexes and transform natural language into optimized queries for context.

With Agent Builder, you can ensure agents receive the most useful context first with controls for relevance and ranking, allowing you to fine-tune scoring, ranking, and filtering logic. Elasticsearch lets you control what matters, why it matters, and how it is prioritized, instead of relying on opaque retrieval behavior. This is all underpinned by Elasticsearch as a scalable data platform to store and scale all your data from text, vectors, metadata, logs, and more on one platform, making it easier to manage context for agents.

Execute complex workflows as reusable tools

While AI agents enable reasoning for complex tasks, much automation depends on reliably executing rules-based actions that enforce specific business logic. Elastic Workflows provides a simple, declarative way to orchestrate internal and external systems to take actions, gather context or data, and integrate them as part of agents. Defined in YAML, workflows are fully composable, allowing them to be as simple or as complex as the job requires. This gives agents an efficient way to take action across the Elasticsearch platform and solutions, as well as with third-party applications.

Integrating a workflow with Agent Builder can be done in three steps (prerequisite: enable workflows with details provided here)

1. Create and save a new workflow using the simple YAML-based editor with built-in autocomplete and testing.

2. Create a new tool in Agent Builder with type “Workflow” and provide a description to help the agent determine when to use the workflow tool.

3. Add the workflow tool to your custom agent.

4. That’s it! Now the agent can call the workflow from within a conversation.

Your agent, your rules

Agent Builder doesn’t lock you into a single development paradigm. Instead, it’s designed to enable open, flexible development approaches for agents with full control of data, relevance, models, interoperability, security and agent design.

Custom agent definitions let you choose exactly which tools an agent can access, embed custom system prompts, tailor the agent’s instructions, and define security boundaries. Agents remain model-agnostic, allowing you to flexibly configure a preferred LLM, both native and across the broader ecosystem, without being locked into a single provider.

Build extensible tools that encapsulate domain-specific logic (e.g., specific index filters, ES|QL joins, analytical pipelines), and constrain them for safe use in production. Full API support enables interoperability with other agentic frameworks, with native support for Model Context Protocol (MCP). A2A integration means you can expose your Elastic agents to other frameworks, services, and client apps, reusing the same data and context engineering logic across integrations.

Agent Builder supports flexible, open development and is designed to integrate easily with popular agent frameworks and platforms. These integrations can be essential to delivering effective agents. As Sam Partee, Co-Founder at Arcade.dev describes,

“Agentic systems fail today because connecting AI to tools and data is complex. Elastic Agent Builder with Arcade.dev gives developers a structured, secure way to handle how agents retrieve context, reason, and act, taking agents from demo to production grade."

Agent Builder also leverages the extensibility of Elasticsearch for handling complex data. As Jerry Liu, CEO at LlamaIndex describes,

“Unlocking enterprise context from unstructured data sources is key to building effective agents. Elastic Agent Builder combined LlamaIndex complex document processing strengthens the critical context layer, helping teams retrieve, process, and prepare data so agents can reason more accurately and deliver better outcomes.”

What can you build?

Agent Builder is already being used for a variety of use cases. Below are a few examples and reference architectures to get started with agents:

• Automate infrastructure: In support scenarios, agents have been used to read, think, and chat, but to date, they cannot reach out and touch the infrastructure they may need to manage. Elastic’s engineering team built an agent for automated infrastructure management as part of a hackathon. The agent actively investigates issues with application infrastructure and takes automated actions. It uses workflows to optimize configurations, respond to issues and scale resources, all based on an intelligent understanding of infrastructure logs.
• Security threat analysis: A security vulnerability agent was developed with Elastic Agent Builder, MCP, and Elasticsearch. It automates threat analysis by correlating internal security data with external threat intelligence. The agent performs semantic search over historical incidents and configurations, augments results with live internet data, and applies LLM reasoning to assess environmental relevance, prioritize risks, and produce actionable remediation. See the reference architecture.
• Technical customer support: Agents can perform multiple support tasks, including case summarization, issue deduplication and creation, and deep technical investigation. Agent Builder enables this with multi-step, hybrid search to find only the most relevant related issues, solutions, and procedures, and formulate root cause hypotheses and remediation plans. Agent Builder can simplify the architecture of complex support systems and accelerate time to delivery.
• Product and content discovery: Agent Builder simplifies the process of exposing complex product catalogs for conversational experiences, while allowing organizations to maintain flexibility to include their own business logic and requirements.
• Build your own: Join the Agent Builder Hackathon, running from January 22 to February 27, 2026. Work with the community to build context-driven, multi-step AI agents that combine search, workflows, tools, and reasoning to automate real-world tasks*

Start building custom agents now

Get started with an Elastic Cloud Trial, and check out the documentation here. For existing customers, Agent Builder is available in Cloud Serverless and on the Enterprise Tier in Elastic Cloud Hosted and self-managed.

* Click here for full terms, conditions, and eligibility requirements for the hackathon