RAG Engine
Turn-key Knowledge Base for AI Agents
A knowledge base for AI agents, built into the AppXen Platform appliance: document parsing, chunking, embeddings on Amazon Titan, and semantic retrieval, exposed as an MCP server. Upload PDFs, DOCX, images, and more. Your agents search your knowledge base with natural language.
What it does
RAG Engine is a knowledge base pipeline for AI agents, built into the appliance. Upload documents through the dashboard or API, and they are automatically parsed, chunked, embedded, and indexed in a vector database. Your agents then search across this knowledge base using natural language, getting back the most relevant passages with citations.
RAG Engine is exposed as a built-in MCP server through MCP Gateway Pro. When your agent calls rag-engine__knowledge_query, it runs a semantic search and returns matching document chunks. There is nothing separate to deploy: it is part of the appliance, so you upload and search.
How It Works
Batch upload with the CLI
The AppXen CLI is the fastest way to load documents. Point it at a directory and it uploads everything, with progress tracking and status polling built in.
$ appxen ingest ./docs/ --recursive Supported Formats
Documents
PDF DOCX HTML Markdown CSV JSON Plain text Images (via OCR)
PNG JPEG TIFF BMP Text extracted via OCR
Key Features
Drag-and-Drop Upload
Upload documents through the dashboard with drag-and-drop or paste text directly. Track ingestion status in real time.
Semantic Search
Vector similarity search powered by pgvector. Adjustable top-k results.
MCP Integration
Exposed as a built-in MCP server with 5 tools. Any agent connected to your gateway can query your knowledge base.
Dashboard
Overview stats, source management, search playground with query history, and ingestion monitoring. All from the console.
Multi-Format Parsing
PDF, DOCX, HTML, Markdown, CSV, JSON, plain text, and images via OCR. Automatic format detection.
Async Ingestion
File uploads are queued and processed in the background. Poll for status or check the dashboard. No request timeouts for large files.
MCP Tools
RAG Engine exposes 5 tools through the gateway. Any AI agent connected to your gateway endpoint can use these tools directly.
knowledge_query Search your knowledge base using natural language. Returns the most relevant document chunks with citations.
knowledge_ingest Add a document to your knowledge base. Provide text directly or base64-encoded file content for binary formats.
knowledge_list_sources List all indexed documents in your knowledge base. Filter by status (queued, processing, ready, failed).
knowledge_delete_source Delete a document and all its chunks from the knowledge base.
knowledge_get_chunk Get a specific chunk by ID for citation follow-up or context expansion.
Example: Search Your Knowledge Base
POST https://<your-appliance-host>/mcp Authorization: Bearer <your API key> { "jsonrpc": "2.0", "method": "tools/call", "params": { "name": "rag-engine__knowledge_query", "arguments": { "query": "What is our refund policy?", "top_k": 3 } } }
Storage vs RAG Engine
MCP Gateway Pro includes both raw file storage and the RAG Engine knowledge base. They serve different purposes:
Storage
Raw file storage for AI agents. Files stay as-is.
- - Agents read files verbatim via MCP tools
- - Good for reference documents agents access directly
- - Images, videos, code files, any format
RAG Engine
Knowledge base pipeline. Documents are parsed, chunked, embedded, and indexed.
- - Agents run semantic search across content
- - Good for large document collections
- - Find relevant info without reading entire files
Pricing
Included in the AppXen Platform appliance: from $0.149 per instance-hour or $949 per year on the large instance types ($0.199 or $1,299 on the xlarge types), billed hourly or annually through AWS Marketplace, plus your own AWS usage and Amazon Bedrock tokens, billed by AWS. AppXen does not meter requests, storage or tokens.
Learn moreSpecs
Dashboard
Run RAG Engine in your AWS account
RAG Engine ships inside the AppXen Platform appliance. It runs in your AWS account on your own Amazon Bedrock, billed hourly or annually through AWS Marketplace, plus your own AWS usage.