Introduction to Grounded Docs MCP Server

Grounded Docs MCP Server solves a critical problem in AI-assisted development: hallucinations and outdated knowledge. When your AI assistant suggests code or answers questions, it often relies on training data that may be months or years old. This leads to deprecated APIs, incorrect syntax, and wasted debugging time. Grounded Docs provides your AI with a personal, always-current documentation index that fetches official docs directly from websites, GitHub repositories, npm packages, PyPI, and local files. Your AI queries the exact version you’re using, grounded in real documentation.

Why Use Grounded Docs?

Eliminate Hallucinations

Ground your AI in real documentation instead of relying on potentially outdated training data

Version-Specific Accuracy

Query documentation for the exact library versions in your project, not generic answers

Privacy First

Runs entirely on your machine - your code and queries never leave your network

Universal Compatibility

Works with any MCP-compatible client: Claude, Cursor, Cline, VS Code extensions, and more

Key Features

Multiple Documentation Sources

Index documentation from any source:

Websites: Official documentation sites (React, Next.js, etc.)
GitHub Repositories: README files, wikis, and markdown docs
Package Registries: npm and PyPI packages with automatic version detection
Local Files: Your team’s internal documentation, project READMEs, and custom guides
Zip Archives: Compressed documentation bundles

Rich File Format Support

The server processes and indexes multiple file types:

Web formats: HTML, Markdown
Documents: PDF, Word (.docx), Excel, PowerPoint
Code: JavaScript, TypeScript, Python, and other source files
Archives: ZIP files with automatic extraction

Intelligent Search

Semantic search is optional but dramatically improves result quality by understanding the meaning of your queries, not just matching keywords.

Choose your search mode:

Keyword search: Fast, no configuration required (default)
Semantic vector search: Understand meaning and context using embeddings from OpenAI, Ollama, Google Gemini, Azure, or AWS Bedrock
Hybrid search: Combine both approaches using Reciprocal Rank Fusion for best results

Flexible Deployment

Run the server in multiple configurations:

Standalone Server
Embedded Mode
Docker
Distributed (Docker Compose)

Single process with web interface and MCP endpoints. Perfect for most users.

npx @arabold/docs-mcp-server@latest

Access the web UI at http://localhost:6280

Runs directly inside your AI assistant with no separate process.

{
  "mcpServers": {
    "docs-mcp-server": {
      "command": "npx",
      "args": ["-y", "@arabold/docs-mcp-server@latest"]
    }
  }
}

Isolated environment with persistent storage.

docker run --rm \
  -v docs-mcp-data:/data \
  -v docs-mcp-config:/config \
  -p 6280:6280 \
  ghcr.io/arabold/docs-mcp-server:latest \
  --protocol http --host 0.0.0.0 --port 6280

Separate worker and coordinator processes for scaling.

docker compose up -d

See Deployment Modes for details.

How It Works

Start the Server

Launch Grounded Docs using npx, Docker, or embedded mode

Add Documentation

Use the web interface or CLI to scrape documentation from URLs or local files

Automatic Processing

The server fetches content, chunks it intelligently, and generates embeddings (if configured)

Connect Your AI

Configure your AI assistant (Claude, Cursor, etc.) to connect to the MCP server

Query & Get Accurate Answers

Your AI now has access to current, version-specific documentation for better responses

Use Cases

Framework Documentation

Keep your AI up-to-date with the latest framework APIs:

npx @arabold/docs-mcp-server@latest scrape react https://react.dev/reference/react

Now ask your AI: “How does the useEffect hook work with cleanup functions?”

Internal Documentation

Index your team’s private documentation:

npx @arabold/docs-mcp-server@latest scrape internal file:///Users/me/company-docs

Your AI can now answer questions about your internal APIs and best practices.

Package-Specific Help

Get help with specific library versions:

npx @arabold/docs-mcp-server@latest scrape lodash npm:[email protected]

Ask: “Show me how to use debounce in lodash 4.17”

Local Project Documentation

Index your project’s README and guides:

npx @arabold/docs-mcp-server@latest scrape my-project file:///Users/me/projects/my-app

Open Source Alternative

Grounded Docs is the open-source alternative to commercial documentation tools:

Context7: Proprietary, cloud-based
Nia: Closed source
Ref.Tools: Limited to web documentation

With Grounded Docs, you get:

Full control over your data
No vendor lock-in
Extensible architecture
Active community development

Architecture Highlights

For a deep dive into the system architecture, see the Architecture documentation.

Content Processing Pipeline:

Fetcher retrieves content from various sources
Middleware transforms HTML/Markdown/PDF to plain text
Semantic splitters chunk content by structure (headers, code blocks)
Greedy optimizer adjusts chunk sizes for embedding quality
Embeddings are generated (optional) and stored in SQLite

Search System:

Vector similarity search using sqlite-vec
Full-text search using SQLite FTS5
Reciprocal Rank Fusion combines results
Configurable ranking weights

Event-Driven Updates:

Real-time progress updates via EventBus
WebSocket subscriptions for distributed mode
Job state persistence for recovery

Next Steps

Quick Start

Get up and running in 5 minutes

Installation Guide

Detailed setup instructions for all deployment modes

Connecting Clients

Configure Claude, Cursor, VS Code, and other AI assistants

Embedding Models

Enable semantic search with OpenAI, Ollama, or other providers

Getting Started

Setup

Guides

Architecture

Infrastructure

Introduction

Introduction to Grounded Docs MCP Server

Why Use Grounded Docs?

Eliminate Hallucinations

Version-Specific Accuracy

Privacy First

Universal Compatibility

Key Features

Multiple Documentation Sources

Rich File Format Support

Intelligent Search

Flexible Deployment

How It Works

Use Cases

Framework Documentation

Internal Documentation

Package-Specific Help

Local Project Documentation

Open Source Alternative

Architecture Highlights

Next Steps

Quick Start

Installation Guide

Connecting Clients

Embedding Models

Build docs developers (and LLMs) love

Getting Started

Setup

Guides

Architecture

Infrastructure

​Introduction to Grounded Docs MCP Server

​Why Use Grounded Docs?

Eliminate Hallucinations

Version-Specific Accuracy

Privacy First

Universal Compatibility

​Key Features

​Multiple Documentation Sources

​Rich File Format Support

​Intelligent Search

​Flexible Deployment

​How It Works

​Use Cases

​Framework Documentation

​Internal Documentation

​Package-Specific Help

​Local Project Documentation

​Open Source Alternative

​Architecture Highlights

​Next Steps

Quick Start

Installation Guide

Connecting Clients

Embedding Models

Build docs developers (and LLMs) love

Introduction to Grounded Docs MCP Server

Why Use Grounded Docs?

Key Features

Multiple Documentation Sources

Rich File Format Support

Intelligent Search

Flexible Deployment

How It Works

Use Cases

Framework Documentation

Internal Documentation

Package-Specific Help

Local Project Documentation

Open Source Alternative

Architecture Highlights

Next Steps