Summary

Cluster work without composing the JSON yourself.

The reach is the useful part: the same session that checked cluster health can read a mapping, add a field, reindex into the new shape and query the result — steps that normally mean four hand-written request bodies.

What it is

A connector that gives an assistant the operational surface of Elasticsearch: it can list what indices exist, read and change their mappings, run a query DSL search, push data in, and manage index templates — with cluster health as the starting point.

What you get

Cluster — `elasticsearch_health` for cluster status, optionally with index-level detail
Indices — `list_indices` with regex support, `create_index` with settings and mappings, `reindex` from one index to another with an optional query and script
Mappings — `get_mappings` for an index, `create_mapping` to create or update the structure
Search and data — `search` with a query DSL body, `bulk` for loading data in
Templates — `create_index_template`, `get_index_template`, `delete_index_template`

Requirements

An Elasticsearch instance and credentials: `ES_HOST` is required, and you must supply either `ES_API_KEY` or both `ES_USERNAME` and `ES_PASSWORD`. `ES_CA_CERT` points at a custom CA certificate for TLS. `ES_HOST` accepts a comma-separated list of node URLs, and the client handles failover and load balancing across them. Runs as `npx -y @awesome-ai/elasticsearch-mcp`. Since the same connection carries create, reindex and delete-template operations, give it credentials scoped to what you actually want an assistant to change.

Setup effort

One command plus a key — npx -y @awesome-ai/elasticsearch-mcp, then supply credentials