Elastic AI Agent

AI agent · Data & analytics agent · Data & analytics by Elastic

Chat with your data and build production-ready AI agents in minutes.

Elastic AI Agent is a customizable conversational assistant for businesses to query and analyze their Elasticsearch data using natural language. It enables automated dashboard creation and incident investigation while grounding responses in enterprise context to ensure high reliability.

Agent specs

The full spec card (context, pricing per token, modalities and more) on AntSpace. Sign up free to see the full spec card.

Category: AI Agents

Deployment
Elastic Cloud (Hosted and Serverless)
Integration Protocols
MCP, A2A, REST API
Query Language Support
ES|QL
Deployment Platform
Elastic Cloud (Hosted and Serverless)
Conversation Limit
30+ turns with accuracy
Token Cost Reduction
Up to 40%
  • Built-in conversational assistant for Elasticsearch data
  • Generates Kibana dashboards from natural language prompts
  • Reduces token costs by up to 40% via context compaction
  • Supports MCP and A2A for external agent collaboration
  • Grounded in enterprise data to prevent hallucinations

Overview

Elastic AI Agent is a built-in conversational assistant designed for businesses to interact with their enterprise data using natural language. It allows users to securely query Elasticsearch indices, perform complex data analysis, and generate functional Kibana dashboards from simple text prompts. The agent is highly customizable through the Elastic Agent Builder, where teams can add domain-specific skills and connect to external tools like messaging services or cloud functions. By utilizing Elasticsearch's hybrid search, it grounds AI responses in the organization's actual data to ensure accuracy and prevent hallucinations. It supports modern protocols like MCP and A2A, enabling it to work as a specialist knowledge agent for external applications like Claude Desktop or Cursor. The platform also manages context automatically, reducing token costs by up to 40% while maintaining accuracy over long conversations.

Data Analysis and Interaction

  • Natural Language Querying: Users can securely query and analyze data stored in Elasticsearch indices using natural language.
  • Automated Dashboard Creation: The agent can turn a simple prompt into a working Kibana dashboard by writing ES|QL and composing visualizations.
  • Customizable Skills: Users can add and create custom, domain-specific skills to improve and shape the agent's behavior for specific tasks.
  • Multi-turn Memory Management: The agent supports multi-turn memory to maintain context across conversations of 30+ turns.

Connectivity and Automation

  • External System Actions: Agents can connect to external systems like messaging services, cloud functions, and third-party APIs to take actions.
  • Secure Credential Management: Credentials for external tools are managed securely through Kibana Stack Connectors.
  • REST API Integration: Custom agents can be exposed via REST API with a single code snippet.

Control and Customization

  • Data Grounding: Responses are grounded in the organization's own data through Elasticsearch's hybrid search to prevent hallucinations.
  • Security and Access Control: Role-based access control (RBAC) and document-level security ensure users only see data they are authorized to access.
  • Refined Instructions: The agent can be customized to follow specific instructions and refined directly within the Agent Builder.

What it does

  • Querying and analyzing data in Elasticsearch indices
  • Generating Kibana dashboards from natural language prompts
  • Investigating incidents and finding root causes
  • Running threat hunts
  • Powering knowledge and support assistants
  • Executing complex workflows as reusable tools
  • Calling external APIs and messaging services

Published results

  • Up to 40% reduction in token costs through automatic context compaction
  • Maintains accuracy across 30+ turns

Use cases

  • Data analysts use natural language to query complex Elasticsearch indices and generate Kibana dashboards instantly.
  • Security teams build agents to investigate incidents, find root causes, and run threat hunts across their data.
  • Support teams deploy agents grounded in their knowledge base to deflect tickets and provide accurate self-service answers.
  • Developers integrate the agent into external apps like Cursor via MCP to provide real-time context from enterprise data.

Specifications

Channels
Kibana UI, MCP-enabled apps (Claude Desktop, Cursor), REST API, A2A (agent-to-agent)
Integrations
Messaging services, Cloud functions, Third-party APIs, Claude Desktop, Cursor, Kibana Stack Connectors
Data sources
Elasticsearch indices, Enterprise data sources via connectors, PDFs, Logs, Product docs
Human control
Users refine instructions directly, manage credentials through Kibana, and can share or extend generated dashboards manually.
Models
Model-agnostic: supports Elastic-managed models (Jina), Anthropic, OpenAI, Amazon Bedrock, Azure AI Studio, and Google AI.
Deployment
Elastic Cloud (Hosted and Serverless)
Languages
Multilingual support via Jina AI models on Elastic Inference Service
Integration Protocols
MCP, A2A, REST API
Query Language Support
ES|QL
Deployment Platform
Elastic Cloud (Hosted and Serverless)
Conversation Limit
30+ turns with accuracy
Token Cost Reduction
Up to 40%
Security Model
RBAC and Document-level security
Model Compatibility
Model-agnostic (Elastic-managed or external)

How it works

  1. Access Agent Builder: Access the Agent Builder within Elastic Cloud to start creating custom agents grounded in your data.
  2. Define Skills and Tools: Add capabilities by creating skills, attaching custom tools, or importing plugins to define how the agent interacts with data and external systems.
  3. Integrate and Deploy: Connect the agent to your Elasticsearch indices and external clients via MCP, A2A, or REST API.

FAQ

What can I build with it?

You can build agents for incident investigation, threat hunting, dashboard creation, and knowledge or support assistants.

Which model providers are supported?

It is model-agnostic, supporting Elastic-managed models through Elastic Inference Service as well as external providers like OpenAI, Anthropic, and Amazon Bedrock.

Does it support MCP and A2A?

Yes, it ships with a built-in MCP server for apps like Claude Desktop and an A2A server for collaborating with external agents.

Start building

Added by AntSpace from public sources (elastic.co).

Related

More AI agents by Elastic