Name: Searchcraft
Author: searchcraft-inc

LABSCO SUMMARY

Search infrastructure set up by conversation.

Standing up an index normally means a schema file, a key, an ingest script and four terminal windows. Here it is a description — the schema, the dataset, the read key — and the same connection then tells you what the index actually holds.

What it is

An administrative surface for a Searchcraft cluster: indexes created and modified, documents ingested and deleted, index statistics read, and the federations, access keys, search and analytics around them.

What you get

An index created with a schema, then updated, patched or deleted
The index list, one index's schema, and statistics for one index or across all of them
Documents added to an index
A document retrieved by ID
Documents deleted by ID, by field, by query, or all of them at once
Federations, access keys, search and analytics reachable over the same connection

Requirements

A recent Node.js, and a Searchcraft cluster to point at with credentials for it. Creating an index that already exists empties it, so check the index list before re-running a setup.

Setup effort

Build from source — git clone https://github.com/searchcraft-inc/searchcraft-mcp-server.git

See installation options

USE CASES

01

Standing up a new search index

Create it with the schema you describe, then ingest the documents in the same session.

create_indexadd_documents

02

Checking what an index actually holds

Read its schema and its statistics rather than assuming the ingest worked.

get_index_schemaget_index_stats

03

Clearing bad data out

Delete by query or by field so a botched import goes without taking the index with it.

delete_documents_by_querydelete_documents_by_field

04

Changing a schema after launch

Patch or update the index definition instead of rebuilding from scratch.

patch_indexupdate_index

05

Auditing a cluster

List every index and read the statistics across all of them in one pass.

list_all_indexesget_all_index_stats

06

Retiring an index for good

Empty it, then delete it, once the app that used it is gone.

delete_all_documentsdelete_index

48 tools

Listed in project docs. Your client may show a different list.

create_index

delete_index

get_all_index_stats

get_index_schema

get_index_stats

list_all_indexes

patch_index

update_index

add_documents

delete_all_documents

delete_document_by_id

delete_documents_by_field

delete_documents_by_query

get_document_by_id

create_federation

BEFORE YOU INSTALL

What to know first

Tools you get

48 tools named in the project's documentation.

Requires

CORE_API_KEY in your environment
ENDPOINT_URL in your environment

INSTALLATION

Connect your client

You clone and build this one yourself.

Server setup

TERMINAL

git clone https://github.com/searchcraft-inc/searchcraft-mcp-server.git
cd searchcraft-mcp-server
yarn
yarn build

Create a .env file at the project's root with ENDPOINT_URL (your Searchcraft cluster endpoint) and CORE_API_KEY — see .env.example.

MCP Clients

Claude Desktop — stdio

CLAUDE DESKTOP — STDIO · JSON

{
  "mcpServers": {
    "searchcraft": {
      "command": "node",
      "args": [
        "<YOUR_PATH_TO_SEARCHCRAFT_MCP_SERVER>/dist/stdio-server.js"
      ],
      "env": {
        "ENDPOINT_URL": "https://<YOUR_CLUSTER>.searchcraft.io",
        "CORE_API_KEY": "<YOUR_CORE_API_KEY>"
      }
    }
  }
}

<YOUR_PATH_TO_SEARCHCRAFT_MCP_SERVER> is the absolute path of the clone you built. The two env values are your Searchcraft cluster endpoint and API key — the same two the Claude Code command passes with --env.

Claude Code

CLAUDE CODE · TERMINAL

claude mcp add searchcraft \
  --env ENDPOINT_URL=https://<YOUR_CLUSTER>.searchcraft.io \
  --env CORE_API_KEY=<YOUR_CORE_API_KEY> \
  -- node <YOUR_PATH_TO_SEARCHCRAFT_MCP_SERVER>/dist/stdio-server.js

Claude Desktop — HTTP

From your built clone, start the HTTP server (it listens on port 3100):

CLAUDE DESKTOP — HTTP · TERMINAL

yarn start

CLAUDE DESKTOP — HTTP · JSON

{
  "mcpServers": {
    "searchcraft": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "http://localhost:3100/mcp"
      ]
    }
  }
}

Open WebUI

From your built clone, start the HTTP server (it listens on port 3100):

OPEN WEBUI · TERMINAL

yarn start

In the Searchcraft pipeline's valves in Open WebUI, set:

OPEN WEBUI · JSON

{
  "MCP_SERVER_URL": "http://localhost:3100/mcp",
  "ENDPOINT_URL": "<YOUR_ENDPOINT_URL>",
  "CORE_API_KEY": "<YOUR_CORE_API_KEY>"
}

Docker

From your clone, build the image first:

DOCKER · TERMINAL

docker build --load -t searchcraft-mcp-server .