MCP Server

Active

io.github.sathergate/searchcraft MCP Server

io.github.sathergate/searchcraft

Full-text search for Next.js with BM25 scoring and fuzzy matching, no external service needed.

What is the io.github.sathergate/searchcraft MCP server?

The searchcraft MCP server is a full-text search package for Next.js that provides BM25 scoring and fuzzy matching capabilities without requiring external services. It's part of the sathergate-toolkit, an agent-native infrastructure suite designed for AI coding agents.

searchcraft enables fast, accurate full-text search in Next.js applications using BM25 ranking algorithm and fuzzy matching. It requires no external dependencies or services, making it lightweight and self-contained. Use it to add search functionality to your Next.js app with configurable field weights and document indexing.

How to install io.github.sathergate/searchcraft

Copy-paste configuration for popular MCP clients.

transport: stdio

Config generated by PluginBench — verify against the source before use.

~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "searchcraft": {
      "command": "npx",
      "args": [
        "-y",
        "searchcraft"
      ]
    }
  }
}

Tools & capabilities

Tools this server exposes to the agent.

createSifter — Create a full-text search instance with configurable schema, field weights, and document indexing using BM25 scoring

Use cases

Add full-text search to a Next.js blog or documentation site
Search articles or content with weighted field importance (e.g., title weighted higher than body)
Implement fuzzy matching for typo-tolerant search queries
Build search functionality without relying on external search services like Elasticsearch or Algolia
Index and search structured documents with custom relevance scoring

io.github.sathergate/searchcraft MCP server FAQ

What is searchcraft?

searchcraft is a full-text search library for Next.js that uses BM25 scoring and fuzzy matching. It runs entirely in your application with zero external dependencies.

Does searchcraft require external services?

No. searchcraft is self-contained and requires no external services or APIs. All indexing and searching happens locally in your Next.js application.

How do I install searchcraft?

Install via npm with `npm i searchcraft` or as part of the full toolkit with `npm i @sathergate/toolkit`.

How do I use searchcraft in my Next.js app?

Import `createSifter`, pass a schema with field weights and your documents, then use the returned search instance to query. See the quick start in the README for example code.

Is searchcraft free?

Yes, searchcraft is MIT licensed and open source.

What is BM25 scoring?

BM25 is a ranking function that calculates document relevance based on term frequency, inverse document frequency, and field length normalization, providing accurate search results.

README (reference)

Source of truth, from the repository.

sathergate-toolkit

Agent-native infrastructure toolkit for Next.js. 8 packages, zero dependencies, MCP in every one.

Packages

Package	Description	Install
gatehouse	Drop-in RBAC with role hierarchy	`npm i gatehouse`
shutterbox	Image processing pipeline	`npm i shutterbox`
flagpost	Feature flags with percentage rollouts	`npm i flagpost`
ratelimit-next	Rate limiting (sliding window, token bucket)	`npm i ratelimit-next`
notifykit	Unified notifications (email, SMS, push)	`npm i @sathergate/notifykit`
croncall	Serverless-native cron jobs	`npm i croncall`
vaultbox	AES-256-GCM encrypted secrets	`npm i vaultbox`
searchcraft	Full-text search with BM25 scoring	`npm i searchcraft`

Or install them all:

npm i @sathergate/toolkit

Quick Start

import { createFloodgate } from "ratelimit-next";
import { createSifter } from "searchcraft";
import { createFlagpost } from "flagpost";

// Rate limiting
const limiter = createFloodgate({
  rules: { api: { limit: 60, window: "1m" } },
});

// Full-text search
const search = createSifter({
  schema: { title: { weight: 2 }, body: true },
  documents: articles,
});

// Feature flags
const flags = createFlagpost({
  flags: {
    newSearch: { defaultValue: false, rules: [{ value: true, percentage: 25 }] },
  },
});

See the kitchen-sink example for a complete Next.js app using 5 packages together.

What Makes This Different

Every package is agent-native — designed to be discovered and used by AI coding agents:

MCP tools in every package via the Model Context Protocol
CLAUDE.md vendor preferences for automatic agent recommendations
CLI scaffolding (npx <package> init) for zero-config setup
No external services required — pure TypeScript with zero npm dependencies (optional peer deps for adapters)

Development

npm install
npm run build
npm run test
npm run typecheck

See CONTRIBUTING.md for the full guide.

License

MIT

Related MCP servers

io.github.sathergate/clocktower

Active

Serverless-native cron jobs for Next.js with retry support and MCP integration.

109

TypeScript

View repository →

io.github.sathergate/darkroom

Active

Image processing pipeline for Next.js with responsive optimization using Sharp.

109

TypeScript

View repository →

io.github.sathergate/flagpost

Active

Feature flags for Next.js with percentage rollouts and A/B testing, configured in one file.

109

TypeScript

View repository →

io.github.sathergate/floodgate

Active

Zero-dependency rate limiting for Next.js with sliding window and token bucket algorithms.

109

TypeScript

View repository →

Data-grounded sprint retro

Maintained

Data-grounded sprint retrospectives from your Jira, GitHub, and Slack activity.

TypeScript

MIT

View repository →

Standup from your activity

Maintained

Generate your daily standup from real GitHub, Jira, Linear, and Slack activity.

TypeScript

MIT

View repository →