Keep your agents in context with Elasticsearch

Turn unstructured business data into trusted context for your LLMs using Elasticsearch, the open, unified platform for context engineering and agentic AI.

Start free trial

The core capabilities of context engineering

Building reliable AI starts with context.

Context engineering connects data, retrieval, tools, and memory — the essential components that enable models to learn, reason, and act from the right information.

What is context engineering?

Connecting the right data
All data, connected across systems and formats, helps reasoning models turn scattered information into meaningful context.
Retrieving what matters most
Like human short-term memory, a model's context is limited. Smart retrieval provides just-in-time context keeping every decision relevant.
Structured understanding and tools
Clear structure and standardized tools help agents interpret context, form accurate conclusions and act with purpose.
Adaptive memory and workflows
Memory and workflows connect past context to future actions, enabling coherent, informed decisions.

The Elastic advantage: Provide your agents with the right information, tools, and guardrails

One datastore for messy enterprise data
The greater the variety of data, the richer your context. Unify unstructured enterprise data such as tickets, logs, documents, and feedback in one datastore with efficient columnar storage that delivers real-time retrieval and trusted context for AI.
Elasticsearch vector database
The best relevance engine for context engineering
Achieve exceptional relevance with hybrid search, semantic reranking, and built-in inference. Elasticsearch gives you the tools to interpret intent, filter by permissions, and rank relevant context so agents retrieve what truly matters.
What is context engineering?
Agent Builder: Where data meets decision
Bring action and control to your context layer. Create custom tools with ES|QL, chat securely with your data, and integrate with external agents. Agent Builder connects your prompts, data, and workflows so you can ship context-driven agents in minutes.
Learn more about Agent Builder
Build safely. Measure everything.
Delivering relevant context spans every layer, from data storage and retrieval to exposure through tools. With built-in telemetry, control, and guardrails, build agents that act safely and stay accurate in production.
Explore LLM observability

All your data. One platform for context.

Engineer your context with Elasticsearch. Add precise search or compose the full conversational AI stack. Start anywhere, with defaults and scaling choices, and shape your relevance journey from simple Q&A to sophisticated agentic workflows.

Launchpad for context engineering

Robust APIs for the lifecycle of your data. Index, search, filter, and apply RBAC — on text, embeddings, geo, timeseries, or metadata.

Ingest and prepare data

Define an ingest pipeline to clean, label, and normalize documents so provenance is clear and fields are parseable.

POST /_ingest/pipeline
{
  "description": "Clean and enrich documents",
  "processors": [
    {
      "set": {
        "field": "source",
        "value": "access_logs_prod"
      }
    },
    {
      "grok": {
        "field": "message",
        "patterns": [
          "{TIMESTAMP_ISO8601:timestamp} User %{WORD:user} accessed - %{IP:ip}"
        ]
      }
    }
  ]
}

Run Inference

Run embeddings on Elastic Inference Service (EIS) — GPU-powered, configuration-free, and built for Elasticsearch.

POST _inference/chat_completion/.rainbow-sprinkles-elastic/_stream
{
  "messages": [
      {
          "role": "user",
          "content": "What is Context Engineering?"
      }
  ]
}

Retrieve and rerank

Run a hybrid query that blends kNN vectors with lexical search, then rescore with a reranker for production-grade relevance. Add filters, geo, or graph as needed without changing the pattern.

Use EIS for Elastic Rerank to access our easy default for reranking, or use a model of your choice.

POST /my-index/_search
{
  "query": {
    "hybrid": {
      "queries": [
        { "knn": { "vector": [ ... ] } },
        { "match": { "text": "hotel gym reviews" } }
      ]
    }
  },
  "rescore": {
    "window_size": 100,
    "query": {
      "rescore_query": { "semantic_rerank": { "model_id": "reranker" } }
    }
  }
}

Secure and govern

Assign granular role-based access controls with document- and field-level security with coverage across widely adopted compliance frameworks.

POST /_security/role/employee_role 
{
  "indices": [
    {
      "names": [ "example-index" ],
      "privileges": [ "read" ],
      "field_security" : {
        "grant" : [ "state", "name.first", "name.last" ]
      }
    }
  ]
}

Evaluate and observe

Continuously test, score, and optimize outputs to keep them accurate and reliable in production.

POST /_ml/evaluate/retrieval
{
  "index": "qa-eval",
  "relevance_field": "labels",
  "metric": { "ndcg": { "k": 10 } }
}

POST /_ingest/pipeline
{
  "description": "Clean and enrich documents",
  "processors": [
    {
      "set": {
        "field": "source",
        "value": "access_logs_prod"
      }
    },
    {
      "grok": {
        "field": "message",
        "patterns": [
          "{TIMESTAMP_ISO8601:timestamp} User %{WORD:user} accessed - %{IP:ip}"
        ]
      }
    }
  ]
}

Assemble your agents

Create and scale tools, agents, and chat interfaces with only a snippet.

Register a reusable tool
Create a custom agent
Converse with your agent

POST /api/agent_builder/tools
{
  "id": "find_client_exposure_to_negative_news",
  "type": "esql",
  "description": "Finds client portfolio exposure to negative news",
  "configuration": { "query": "ES|QL query here" },
  "params": { "time_duration": { "type": "keyword" } }
}

Best in class? Built right in

Native integrations to all the leading AI products — so your apps go further, faster

Rightsize your relevance journey

Elasticsearch gives you relevance control at every level — from precise search to the full conversational AI stack.

Explore the full tuning journey in our blog on Elasticsearch Labs.

Start free trial

Start with relevance
Store structured and unstructured data. Use hybrid search to retrieve accurate, meaningful context from across your sources.
Engineer your context
Use ELSER or multilingual embedding models with hybrid retrieval and rerankers to deliver precise, domain-specific context for your agents.
Orchestrate your agents
Use Agent Builder to connect tools, define workflows, and create reliable, context-driven agents that reason and act on your data.

Frequently asked questions

Why does context matter for AI and agents?

Agents struggle to maintain long-term context, state, and memory across workflows. Context is what keeps them coherent, aware, and consistent over time. Without it, even powerful models lose track of what matters, leading to gaps, hallucinations, or misinterpretations. Context engineering keeps every response grounded in accurate, relevant, and timely information.

Keep your agents in context with Elasticsearch

The core capabilities of context engineering

Connecting the right data

Retrieving what matters most

Structured understanding and tools

Adaptive memory and workflows

The Elastic advantage: Provide your agents with the right information, tools, and guardrails

One datastore for messy enterprise data

The best relevance engine for context engineering

Agent Builder: Where data meets decision

Build safely. Measure everything.

All your data. One platform for context.

Launchpad for context engineering

Ingest and prepare data

Run Inference

Retrieve and rerank

Secure and govern

Evaluate and observe

Assemble your agents

Register a reusable tool

Create a custom agent

Converse with your agent

Best in class? Built right in

Rightsize your relevance journey

Start with relevance

Engineer your context

Orchestrate your agents

Frequently asked questions

Why does context matter for AI and agents?

How does Elasticsearch enable context engineering?

Can I use my own models or frameworks?