How much does Elastic cost?

Elastic Cloud Serverless bills by usage: Agent Builder at $0.025 an execution after 1,000 free, Workflows at $0.0108 an execution after 10,000 free, the Elastic Managed LLM at $4.50 per million input tokens and $21 per million output tokens, and compute from $0.09 to $0.14 per VCU hour plus storage. Elastic Cloud Hosted sells Platinum and Enterprise subscriptions, a free trial is available, and self managed deployments need the matching subscription.

Does Elastic support MCP and tool calling?

Full under the Agentic Index verification standard. Elastic publishes public evidence of Model Context Protocol (MCP) support, tool calling, and API or SDK extensibility for connecting agents to external tools. Evidence basis: Kibana REST APIs create and manage agents, tools and conversations and run agent turns, with an API tutorial and scoped API keys for clients. An MCP server exposes Agent Builder tools to Claude Desktop, Cursor and LangChain apps, and an A2A. Graded from public evidence only against the Agentic Index 14 point capability taxonomy. No vendor pays for placement.

Does Elastic have SOC 2, SSO and role based access control?

Full under the Agentic Index verification standard. Elastic publishes public evidence of enterprise security and identity controls, covering a compliance posture such as SOC 2 or ISO 27001, single sign on (SSO or SAML), and role based access control (RBAC). Evidence basis: Three layers decide what a user or client can do with agents. Kibana feature privileges govern Agent Builder, connector and workflow operations, Elasticsearch cluster and index privileges govern which data agents and tools can query, and Ki. Graded from public evidence only against the Agentic Index 14 point capability taxonomy. No vendor pays for placement.

Can Elastic be self hosted or run in a private VPC?

Full under the Agentic Index verification standard. Elastic publishes public evidence of self hosted or private VPC deployment, along with data residency options for controlling where agent data is processed and stored. Evidence basis: Serverless projects run in a region the customer picks from AWS, Microsoft Azure and Google Cloud regions when creating the project, and the region cannot change afterward. Elastic Cloud Hosted offers its own region list, and the same Agent. Graded from public evidence only against the Agentic Index 14 point capability taxonomy. No vendor pays for placement.

Back to vendors

Elastic

Also known as: Elastic Agent Builder, Elastic Workflows, Elasticsearch, elastic.co

Visit site

Agent builderindependentVerified 2026-10-18

Entry pricePay as you go on Elastic Cloud Serverless. Agent Builder $0.025 per execution after 1,000 free, Workflows $0.0108 per execution after 10,000 free, Elastic Managed LLM $4.50 per million input and $21 per million output tokensFull pricing detail

Elastic, the company behind Elasticsearch, sells Agent Builder for creating AI agents that search, reason over and act on everything stored in Elasticsearch, and Elastic Workflows for automation that calls those agents. Agents use built in and custom tools, external MCP servers and workflows, and reach Claude, Cursor or other agents through MCP and A2A.

Elastic sells Elasticsearch and Kibana as the base of three solutions, Search, Observability and Security, delivered as Elastic Cloud Serverless, Elastic Cloud Hosted or self managed software. Its agent layer sits on top of whatever data a customer already keeps in Elasticsearch, from logs and metrics to security events and product catalogs.

Elastic Agent Builder, generally available since version 9.3, lets teams chat with their data through the built in Elastic AI Agent or build custom agents with their own instructions, tools and skills. Tools query indices with hybrid search or ES|QL, call workflows, or come from external MCP servers, and built in skills such as threat hunting package tools and query templates for a domain.

Elastic Workflows, generally available since 9.4, runs YAML defined automations on manual, scheduled or alert triggers, with Elasticsearch, Kibana, security and external connector steps, an ai.agent step that hands a task to an agent, and waitForApproval steps that pause for a person. Security teams use the same engine as native SOAR for alert triage, enrichment and response. Agents and tools are exposed to outside clients through Kibana APIs, an MCP server and an A2A server.

On Elastic Cloud Serverless, Agent Builder costs $0.025 an execution after 1,000 free, Workflows $0.0108 an execution after 10,000 free, and the Elastic Managed LLM $4.50 per million input tokens and $21 per million output tokens, on top of compute and storage. Access runs through Kibana feature privileges, Elasticsearch index privileges, spaces and API keys, and Elastic Cloud Serverless holds SOC 2 Type 2, ISO 27001, 27017 and 27018, PCI DSS, HIPAA and CSA STAR. Projects run in a chosen AWS, Azure or Google Cloud region, and self managed clusters bring their own model or connect to Elastic's inference service.

Elastic fits organizations that already run search, observability or security on Elasticsearch and want agents and automation where that data lives, with fine grained access control and traces. It fits less well for teams looking for agents that keep a learned memory of users, operate browsers, or come with a testing suite for agent changes, and its nightshift AI SRE agents are still in early access.

Vendor details

Canonical URL

https://www.elastic.co

Subcategory

Agent builder and workflow engine over Elasticsearch data, with agents exposed through MCP and A2A and native SOAR for security teams

Funding status

Public company listed on the New York Stock Exchange (ESTC), the maker of Elasticsearch and Kibana.

Company status

independent

Use cases & customers

Primary use cases

custom agents grounded on Elasticsearch indicessecurity alert triage, enrichment and response with native SOARscheduled and alert driven automation with approval gatesexposing Elasticsearch agents and tools to Claude, Cursor and other agentsconversational analysis of logs, metrics and business data

Target customers

companies already running Elasticsearch for search, logs or securitysecurity operations teams replacing standalone SOARplatform and SRE teams automating incident triagedevelopers building retrieval heavy agents on their own data

Deployment options

Elastic Cloud Serverless on AWS, Azure or Google CloudElastic Cloud HostedSelf managed, Elastic Cloud on Kubernetes or Elastic Cloud Enterprise

Integrations

Agents query Elasticsearch indices with hybrid and semantic search and ES|QL, import tools from external MCP servers, and call Elastic Workflows, whose steps cover Elasticsearch, Kibana, Elastic Security and external systems through Kibana connectors. Models come from the Elastic Managed LLM or inference endpoints for OpenAI, Anthropic Claude, Google Gemini and others. Agents and tools are exposed to Claude Desktop, Cursor and LangChain apps through the MCP server, to other agents through the A2A server, and through Kibana REST APIs.

In practice

A security team routes every high severity detection alert into a workflow. An ai.agent step investigates the related events with the threat hunting skill, the workflow enriches the host and user, and a waitForApproval step asks an analyst before it isolates the endpoint.

An ecommerce company builds a custom agent with index search tools over its product catalog and order data. Support engineers ask it in Kibana why an order stalled, and the agent queries the logs with ES|QL and returns the failing service and time window.

A platform team connects Claude Desktop to the Agent Builder MCP server, so engineers can query production logs and trigger a cache flush workflow from their own assistant under the permissions of a scoped API key.

Sources & related URLs

Related / legacy domains

https://www.elastic.co/docs/explore-analyze/ai-features/elastic-agent-builder https://www.elastic.co/docs/explore-analyze/workflows https://www.elastic.co/security/soar https://www.elastic.co/trust

Research sources

Agentic Index coverage score

11.5 / 14 capabilities · 82%

Integrations & Tool Calling	Full
Agents pick from built in tools, custom index search and ES\|QL tools, tools imported from external MCP servers, and workflow tools that run an Elastic Workflow from a conversation. Workflows reach outside systems through Kibana connectors as external action steps, alongside authenticated Elasticsearch, Kibana and named security steps that change alert status, set tags or call any Kibana API. That gives an agent a path to act in ticketing, chat and paging systems as well as inside Elastic, with every tool carrying a typed signature and structured output the agent parses. Sourceelastic.co/docs/explore-analyze/ai-features/agent-builder/toolsread 2026-10-18
Workflow Orchestration	Full
Elastic Workflows runs versioned YAML automations with steps, inputs, context variables, error handling and composition of one workflow inside another, and an ai.agent step hands a task to an Agent Builder agent in the middle of a deterministic process. Agents can call workflows as tools, workflows can be created by describing them in Agent Chat, and agents can run sub agents in the foreground or background. Security teams use the same engine as native SOAR, where playbooks handle enrichment, escalation and response and agents reason through the cases that go off script. Sourceelastic.co/docs/explore-analyze/ai-features/agent-builder/agents-and-workflowsread 2026-10-18
Knowledge Grounding & RAG	Full
Grounding is the core of the product. Agents answer from the customer's own Elasticsearch indices, the same store that holds its logs, security events, documents and catalogs, using hybrid lexical and vector search, semantic search with Jina embeddings and rerankers, and ES\|QL queries. Index search tools can target remote clusters through cross cluster search, and context engineering primitives decide which data and tools reach the model on each turn. Because the indices are the system of record, new documents and events are searchable by agents as soon as they are ingested, at whatever scale the cluster holds. Sourceelastic.co/docs/explore-analyze/ai-features/elastic-agent-builderread 2026-10-18
Human Oversight & Guardrails	Full
Workflows pause on waitForInput or waitForApproval steps, present structured findings to a responder, and resume with that person's decision. Elastic's guidance puts these gates in front of isolating a host, blocking a user or deleting data, at escalations that page an on call engineer, and on new automations that ask for approval on each action for their first runs. Inside Agent Builder conversations, human in the loop prompts ask the user before the agent continues, though a sub agent cannot collect those answers. Sourceelastic.co/docs/explore-analyze/workflows/authoring-techniques/human-in-the-loopread 2026-10-18
Security, Identity & Governance	Full
Three layers decide what a user or client can do with agents. Kibana feature privileges govern Agent Builder, connector and workflow operations, Elasticsearch cluster and index privileges govern which data agents and tools can query, and Kibana spaces scope agents, tools and conversations, all assigned through roles for people or API keys for programmatic clients with least privilege examples. Elastic runs an ISO 27001 information security system including 27017 and 27018, and Elastic Cloud Serverless on AWS is audited or certified under SOC 2 Type 2, PCI DSS, HIPAA and CSA STAR. Sourceelastic.co/docs/explore-analyze/ai-features/agent-builder/permissionsread 2026-10-18
Observability & Auditability	Full
Agent Builder writes execution traces into the customer's own Elasticsearch deployment, and a prebuilt traces dashboard shows token usage and cost by model and provider, conversation and agent latency, which agents run most and where tool calls fail, per Kibana space. In Agent Chat a user can open each reasoning step and tool call behind an answer, and trace data can feed alerts like any other index. Workflow runs are monitored with their own execution history, and token usage per conversation is tracked for billing. Sourceelastic.co/docs/explore-analyze/ai-features/agent-builder/agent-traces-dashboardread 2026-10-18
Memory & State Persistence	Partial
Agent Chat keeps a conversation history that people reopen, rename or delete through the UI or API, and agents themselves are persisted, space aware objects. Within a conversation the agent carries its earlier turns forward. No store holds what an agent learns about a user or a case from one conversation to the next. The investigation memory that learns from past incidents belongs to the nightshift AI SRE agents, which are in early access. Sourceelastic.co/docs/explore-analyze/ai-features/agent-builder/chatread 2026-10-18
Deployment & Data Residency	Full
Serverless projects run in a region the customer picks from AWS, Microsoft Azure and Google Cloud regions when creating the project, and the region cannot change afterward. Elastic Cloud Hosted offers its own region list, and the same Agent Builder runs on self managed clusters, Elastic Cloud on Kubernetes and Elastic Cloud Enterprise. Those deployments bring their own model through an inference endpoint or connect to Elastic's inference service, so agents can run entirely inside the customer's infrastructure. Sourceelastic.co/docs/deploy-manage/deploy/elastic-cloud/regionsread 2026-10-18
Prebuilt Agents / Templates / Packs	Full
The Elastic AI Agent ships ready in every Kibana space, and built in skills turn it into a specialist: the threat hunting skill, for instance, brings its own tool set and query templates and replaces the standalone Threat Hunting Agent of earlier versions. Built in tools cover search, mapping inspection and ES\|QL out of the box and cannot be altered. Workflows add managed workflows and named alert triage and attack triage steps for security teams, so a buyer starts from packaged capability and adds custom agents only where needed. Sourceelastic.co/docs/explore-analyze/ai-features/agent-builder/builtin-agents-referenceread 2026-10-18
Triggers & Channel Coverage	Full
Workflows start on manual triggers from the UI or API, on schedules set as intervals of minutes, hours or days, or on alert triggers that fire when a detection or alerting rule raises an alert and carry the alert's metadata, entities and source events into the run. Because a workflow can include an ai.agent step, those triggers start agent work without anyone opening a chat. Agents are also reached interactively in Agent Chat in Kibana and from outside clients through the MCP server, the A2A server and the REST API. Sourceelastic.co/docs/explore-analyze/workflows/triggersread 2026-10-18
Model Flexibility & Routing	Full
Elastic Cloud deployments start with Elastic Managed LLMs on the Elastic Inference Service and need no API keys. Customers add their own models as inference endpoints for providers such as OpenAI, Anthropic Claude and Google Gemini, and a model selector in Agent Chat switches between every configured model, with a single request able to use a different one. Self managed deployments include no preconfigured model and choose between their own endpoint and Elastic's service through Cloud Connect. Sourceelastic.co/docs/explore-analyze/ai-features/agent-builder/modelsread 2026-10-18
APIs / SDKs / MCP Extensibility	Full
Kibana REST APIs create and manage agents, tools and conversations and run agent turns, with an API tutorial and scoped API keys for clients. An MCP server exposes Agent Builder tools to Claude Desktop, Cursor and LangChain apps, and an A2A server lets other agents call Elastic agents through agent card and message endpoints authenticated by API key, with synchronous message/send and no streaming yet. Workflows have their own API for manual runs, and plugins extend Agent Builder further. Sourceelastic.co/docs/explore-analyze/ai-features/agent-builder/a2a-serverread 2026-10-18
Testing, Debugging & Optimization	Not documented
Agent Builder has no evaluation suite: no saved test questions, expected answers or scores to compare one version of an agent, its instructions or its model against another. Teams judge changes by reading traces, reasoning steps and the traces dashboard after the fact, and a new workflow can run in an approval on every action mode for its first runs before it is trusted to act alone. Sourceelastic.co/docs/explore-analyze/ai-features/agent-builder/agent-traces-dashboardread 2026-10-18
Browser / Computer-use	Not documented
Agents work through search, ES\|QL, workflow and MCP tools against Elasticsearch, Kibana and the systems reachable through Kibana connectors. There is no hosted browser, desktop session or screen control among the built in tools, and an action in an application without an API would have to come from an external MCP server the customer connects. Elastic's synthetic monitoring does script real browsers, but it checks the uptime and journeys of a customer's web pages for the observability product, and agents do not drive it. An agent that needs to act on a website reaches it only through whatever API or connector exists. Sourceelastic.co/docs/explore-analyze/ai-features/agent-builder/toolsread 2026-10-18

The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded

Pricing

Pay as you go on Elastic Cloud Serverless. Agent Builder $0.025 per execution after 1,000 free, Workflows $0.0108 per execution after 10,000 free, Elastic Managed LLM $4.50 per million input and $21 per million output tokens

Per Agent Builder execution, per workflow execution and per million LLM tokens, on top of ingest, search and machine learning compute per VCU hour, storage per GB month and egress

Trial available

Included quota

1,000 Agent Builder executions and 10,000 workflow executions free on Serverless, and 50 GB of free storage on vector profiles.

What is public

Serverless unit prices for Agent Builder and Workflows executions, managed LLM tokens, Elastic Inference Service tokens, ingest, search and machine learning compute, storage, egress and cross project search, with the free allowances.

Billing mechanics

Sign up for Elastic Cloud and pay as you go for Serverless usage, or buy Hosted and self managed subscriptions, with contracts and committed spend through sales.

Cost watchouts

An agentic turn uses more tokens than the visible conversation, since tool calls, results and reasoning all pass through the model. Ingest is metered on uncompressed, enriched volume, which runs well above raw data size. Bring your own model shifts token cost to the provider.

Variable cost rationale

Agent executions, workflow runs and model tokens each meter separately and stack on the compute and storage that hold the data agents search, so cost follows both agent traffic and data volume.

Additional watchouts

Agent work meters three ways at once, per execution, per token and per compute hour, so a busy agent costs more than its execution count suggests.

Overage / add-ons

Pure usage billing past the free execution allowances, with no plan ceiling on Serverless.

Sales call required

Mixed (some tiers require a call)

Free / trial

Free trial of Elastic Cloud, plus 1,000 Agent Builder executions and 10,000 workflow executions free on Serverless

Lowest paid plan

No plan floor on Serverless; usage is metered from the first unit past the free allowances. Elastic Cloud Hosted is sold as Platinum and Enterprise subscriptions.

Commercial notes

Elastic is a public company. Elastic Cloud can be bought through the AWS, Azure and Google Cloud marketplaces, and Platinum and Enterprise subscriptions carry a 99.95% monthly uptime SLA.

Key ambiguities

Execution prices are stated as starting rates, and the Workflows and Agent Builder prices took effect on 1 May 2026.

Missing data

Hosted and self managed subscription prices for the tiers that include Agent Builder.

Agentic Index verified 2026-10-18

Official pricing page Docs

Data confidence: high