Summary

The model can read the mapping before it writes the query.

Query DSL is unforgiving about field names and types, and that is exactly what the schema resource hands over. Search, index management and the mapping arrive together, so the query gets written against what the cluster actually has.

What it is

An MCP server for Elasticsearch clusters. Four tools cover searching and index management, and every index also appears as a resource at `elasticsearch://<host>/<index>/schema` carrying its field names, types and mapping — discovered from cluster metadata rather than declared by hand.

What you get

A search executed against an index with raw Elasticsearch query DSL, returning the hits
Every index in the cluster listed, with its information
A new index created, with mappings and settings if you supply them
A document indexed, with an ID if you choose one
The mapping for any index exposed as a resource, so a model can read the schema before writing a query against it

Requirements

An Elasticsearch endpoint the server can reach — it is passed as an argument in the client config, not as an environment variable. The README's own quick start runs a single-node cluster with `xpack.security.enabled=false`, so if your cluster has security switched on, work out that path before pointing this at it. MIT licensed.