Bucketeer Docs Local MCP Server

by bucketeer-io

1 stars
293 downloads
Not rated
GitHub

About

A local server to query Bucketeer documentation, which automatically fetches and caches content from its GitHub repository.

Details

Author
bucketeer-io
GitHub stars
1
Downloads
293
Categories
Search, Other, Knowledge Base

- Automatically fetches .mdx files from the GitHub repository.
- Intelligent indexing with keyword extraction and relevance scoring.
- Caches documents locally using SHA‑based change detection.
- Two exposed tools: search_docs and get_document.
- Configurable default search result limit (default: 5).
- Supports recursive directory traversal for all documentation subdirectories.

Setting up with Highlight

This MCP is not yet compatible with Highlight’s one-click setup. However, you can still use it with Highlight by following these steps:

  1. Download and install Highlight from highlightai.com/download
  2. Navigate to the plugins tab and select "Add Custom Plugin"
  3. Configure the plugin with the settings below
    Plugin Name Bucketeer Docs Local MCP Server
    Command (node, npx, python, etc.)

    Please refer to the README for specific instructions on how to obtain API keys or other required environment variables.

  4. Enable "Start Automatically" if you want the plugin to start when Highlight launches

From the repository

Install dependencies (npm install), build the project (npm run build), and build the document index (npm run build:index). Start the server with npm start or run it via npx @bucketeer/docs-local-mcp-server. Configure the server in your MCP client (e.g., Cursor or Claude Desktop) by adding the npx command to its mcpServers configuration or by using the provided Cursor install deeplink.

Claude Desktop / Cursor

Paste into your MCP client config file to install this server.

{
    "mcpServers": {
        "bucketeer docs local mcp server": {
            "bucketeer-docs-local-mcp-server": {
                "command": "npx",
                "args": [
                    "@bucketeer/docs-local-mcp-server",
                    "build-index"
                ]
            }
        }
    }
}

McpServers

{
    "bucketeer-docs-local-mcp-server": {
        "command": "npx",
        "args": [
            "@bucketeer/docs-local-mcp-server",
            "build-index"
        ]
    }
}

Bucketeer Docs Local MCP Server

Overview

This project provides a Model Context Protocol (MCP) server for Bucketeer documentation. It offers an interface for searching and retrieving content from Bucketeer's feature flag and experimentation platform documentation, enabling AI assistants to provide accurate information about Bucketeer's features and usage.

Environment Setup

Requirements

- Node.js 18+
- npm

Installation Steps

1. Clone the repository:

git clone <repository-url>
cd bucketeer-docs-local-mcp-server

2. Install dependencies:

npm install

3. Build the project:

npm run build

4. Build the document index:

npm run build:index

Starting the Server

npm start

Document Sources

The server automatically fetches and indexes documentation from the bucketeer-io/bucketeer-docs repository:

- GitHub Repository Integration:
- Automatically fetches .mdx files from the docs/ directory and all subdirectories
- Processes frontmatter and markdown content for optimal search indexing
- Caches fetched content using SHA hashes and only updates when files are modified
- Supports recursive directory traversal to capture all documentation files

- Intelligent Indexing:
- Extracts keywords from titles, descriptions, headers, and content
- Builds searchable index with relevance scoring based on keyword matches and full-text search
- Optimized for Bucketeer-specific terminology (feature flags, experiments, SDKs, targeting, etc.)
- Handles frontmatter extraction (title, description) from MDX files

- Cache Management:
- Files are cached locally in files/docs/ directory as JSON files
- Document index is stored in files/index/document-index.json
- GitHub cache stored in files/docs/github_cache.json with SHA-based change detection
- Use npm run build:index:force to force rebuild the entire index

Using with npx

First-time Setup

1. Build the document index:

npx @bucketeer/docs-local-mcp-server build-index

2. Use in your MCP configuration as shown in the next section.

Updating the Index

To update the documentation index (e.g., when new documentation is available):

npx @bucketeer/docs-local-mcp-server build-index --force

Cursor and Claude Desktop Configuration

Configure the MCP Server by adding the following to your mcp.json or claude_desktop_config.json file, referring to the documentation for Cursor (https://docs.cursor.com/context/model-context-protocol#configuring-mcp-servers) and Claude Desktop (https://modelcontextprotocol.io/quickstart/user):

Quick Install with Cursor Deeplink

For Cursor users, you can install the MCP server with a single click using the deeplink below:

Install MCP Server

This will automatically configure the MCP server in your Cursor settings. After clicking the link, Cursor will prompt you to install the server.

Option 1: Using npx (Recommended)

{
  "mcpServers": {
    "bucketeer-docs": {
      "type": "stdio",
      "command": "npx",
      "args": ["@bucketeer/docs-local-mcp-server"]
    }
  }
}

Option 2: Using local installation

{
  "mcpServers": {
    "bucketeer-docs": {
      "type": "stdio",
      "command": "npm",
      "args": ["start", "--prefix", "/path/to/bucketeer-docs-local-mcp-server"]
    }
  }
}

Usage

When the MCP server is running, the following tools are available:

1. search_docs - Search Bucketeer Documentation

- Parameter: query (string) - The search query - Parameter: limit (number, optional) - Maximum number of results to return (default: 5)

Example:

{
"name": "search_docs",
"arguments": {
"query": "feature flags SDK integration",
"limit": 5
}
}

Response: Returns an array of search results with title, URL, path, description, excerpt, and relevance score.

2. get_document - Get Specific Document Content

- Parameter: path (string) - Document path obtained from search results

Example:

{
"name": "get_document",
"arguments": {
"path": "getting-started/create-feature-flag"
}
}

Response: Returns the full document content including title, description, URL, and complete markdown content.

Development Commands

- npm run build - Compile TypeScript files to dist/ directory
- npm run build:index - Build/update the document index from GitHub repository
- npm run build:index:force - Force rebuild the entire index (ignores cache)
- npx @bucketeer/docs-local-mcp-server build-index - Build index using npx
- npx @bucketeer/docs-local-mcp-server build-index --force - Force rebuild index using npx
- npm run dev:index - Build and update index in development mode
- npm run dev - Build and start server in development mode
- npm run lint - Run Biome linting
- npm run lint:fix - Run Biome linting and fix linting errors

Configuration

The server is configured via src/config/index.ts:

- siteName: "Bucketeer"
- websiteUrl: "https://docs.bucketeer.io"
- githubRepo: "https://github.com/bucketeer-io/bucketeer-docs"
- docsDirectory: "docs" (directory in GitHub repo containing documentation)
- searchLimitDefault: 5 (default number of search results)
- useGithubSource: true (always uses GitHub as source)

File Structure

files/
├── docs/           # Cached JSON files from GitHub repository
├── index/          # Document search index
│   └── document-index.json
└── [created automatically when building index]

Architecture

The server consists of several key components:

1. GithubDocumentFetcher: Recursively fetches .mdx files from the GitHub repository
2. IndexManager: Builds and manages the searchable document index
3. SearchService: Provides search functionality with keyword matching and full-text search
4. MCP Server: Exposes tools via the Model Context Protocol

License

Apache License 2.0, see LICENSE.

Contributing

We would ❤️ for you to contribute to Bucketeer and help improve it! Anyone can use and enjoy it!

Please follow our contribution guide here.

No reviews yet — be the first

Sign in to leave a review

Use Google, GitHub, or an email account so ratings stay tied to real people.

Email sign in

No reviews posted yet.