Industry News - October 22, 2025
Elastic introduces Agent Builder to simplify AI agent development
Elastic announced Agent Builder, a set of capabilities built on Elasticsearch that lets developers build custom AI agents on company data in minutes. It also ships with a native conversational interface so teams can explore, analyze, and optimize any dataset already living in Elasticsearch.
Agents depend on context. Most enterprises scatter that context across documents, emails, business apps, tickets, and customer feedback. That's why context engineering matters: getting the right data into the agent at the right time. Agent Builder pulls development, configuration, execution, customization, and observability into one place-inside Elasticsearch-so you spend less time wiring systems and more time shipping useful agents.
Elasticsearch already sits at the center of many search and retrieval workflows. Agent Builder builds on that foundation to make relevance, guardrails, and observability part of the default experience. Learn more about Elasticsearch here: Elasticsearch.
What you can do with Agent Builder
- Immediately chat with company data: Use the built-in conversational agent to ask questions across your Elasticsearch indexes. Turn static data into an interactive interface without writing glue code.
- Increase relevance with intelligent tools: Built-in tooling selects the right indexes, understands schema, and translates natural language into optimized semantic, hybrid, or structured queries-returning only the most relevant context to the LLM.
- Build custom tools with ES|QL: Define precise tools using ES|QL to control which data is retrieved, how it's filtered, and what gets sent to the model. Gain fine-grained control over relevance, accuracy, and data access.
- Create custom agents: Set a system prompt, choose allowed tools, and apply a specific security profile. Ship agents that match your use case and compliance needs.
- Integrate via MCP and A2A, safely: Connect external agents and applications through MCP and A2A, while keeping governance and execution in the Elasticsearch layer.
Why it matters for engineering teams
Most agent projects stall on context: which index to search, which fields to trust, and how to keep queries consistent under load. Agent Builder bakes these choices into the platform with clear levers for search, retrieval, and oversight. That reduces custom middleware, cuts failure points, and shortens the path from proof-of-concept to production.
How it works in practice
- Point Agent Builder at your existing Elasticsearch indexes and configure basic access.
- Ask a natural language question; the agent identifies relevant indexes and picks the right query strategy (semantic, hybrid, or structured).
- Add ES|QL-based tools to enforce filters, join datasets, or gate sensitive fields before anything reaches the model.
- Set agent parameters (system prompt, tool access, security profile), then test in the conversational UI.
- Use built-in observability to track prompts, tool calls, query performance, and response quality. Iterate fast.
Where this fits in your stack
- Support and Ops: Triage tickets, summarize threads, and surface known fixes from knowledge bases and runbooks.
- Analytics self-serve: Let stakeholders ask questions in plain English and get validated queries and sourced answers.
- Developer knowledge: Query code search, design docs, and incident reports with consistent, auditable retrieval.
- Compliance by default: Keep execution inside Elasticsearch, apply role-aware access, and log what the agent saw and did.
What to try first
- Enable the conversational agent on your most complete index and run 10 representative queries from real users.
- Create one ES|QL tool that enforces key filters (time window, team, environment) to improve precision.
- Define a minimal system prompt plus two allowed tools; ship a scoped pilot to one team.
- Review observability data, adjust retrieval settings, and lock in guardrails before expanding access.
If you're upskilling your team on agent patterns and retrieval, explore role-based AI training for developers here: Complete AI Training - Courses by job.
Your membership also unlocks:
