Loading

Get started with Elastic Agent Builder

To start using Elastic Agent Builder you need to deploy Elasticsearch and Kibana.

If you don't already have an Elastic deployment, you can start a free trial or refer to Deploy > Choosing your deployment type.

Admonition

This feature requires the appropriate Elastic Stack subscription or Serverless project feature tier.

  1. Access Elastic Agent Builder

    On non-serverless deployments, Elastic Agent Builder availability depends on the navigation mode of your Kibana space:

    You can also search for Agents in the global search field.

    Elastic Agent Builder is the default chat experience in all serverless project types.

    You can also search for Agents in the global search field.

  2. Note

    To learn about required privileges for Elastic Agent Builder, refer to Permissions and access control.

  3. Ingest data into Elasticsearch

    If you already have data in Elasticsearch, skip ahead to start a conversation.

    If you need to add data first, choose the path that fits:

  4. Start a conversation

    The Agent Chat UI provides a conversational interface where you can interact with agents and explore your data using natural language. The default Elastic AI Agent is ready to use immediately.

    Use chat to ask questions, request data analysis, try built-in skills, or create dashboards and visualizations. You can also invoke a specific skill with a slash command, inspect the agent's reasoning and tool calls, and confirm any proposed write actions before changes are applied.

    Learn more in Agent Chat.

  5. Configure model (optional)

    On Elastic Cloud Hosted and Elastic Cloud Serverless, Elastic Agent Builder comes with preconfigured models ready to use. To review recommended models, switch models, or add your own, refer to model selection and configuration.

  6. Build skills, tools, and agents

    After you test the default Elastic AI Agent, create a custom skill for a specific workflow. Skills package task-specific instructions, context, and the tools needed to complete the workflow.

    Add tools when the skill needs to retrieve data, run queries, call APIs, or take action. Create a custom agent when you need a distinct persona, system prompt, model configuration, or set of enabled skills. You don't need a separate agent for every workflow: a single agent can use skill descriptions to choose the right skills and tools for the user's request.

    To build programmatically, try the Elastic Agent Builder API tutorial or explore the Kibana APIs.

Tip

For solution-specific guidance, refer to Agent Builder for Observability and Agent Builder for Elastic Security.