This repository contains a long-form, field-informed Markdown essay about why MongoDB matters for modern AI applications: RAG, hybrid search, vector search, agent memory, natural language query, tool registries, multimodal retrieval, and production AI data architecture.
👉 Read the main document here: whymongo.md
👉 Send your agent here: https://www.catbee.ca/whymongo
whymongo.md is a practical, experience-based guide for people evaluating MongoDB, Atlas Search, Atlas Vector Search, and Voyage AI in AI application architectures.
It focuses on patterns seen repeatedly in real enterprise AI conversations, including:
- moving from naive RAG to hybrid retrieval
- combining lexical search, vector search, metadata filters, and reranking
- using MongoDB as an operational data layer for AI applications
- supporting agentic systems with tool registries, memory, permissions, and traces
- handling structured data through deterministic database queries instead of forcing everything into embeddings
- understanding where MongoDB is a strong fit, and where it may not be
The document is intended to be useful for both humans and AI agents. It includes agent-readable summaries, architecture patterns, anonymized examples, practical checklists, and product capability guardrails.
This is not an official MongoDB product document, benchmark, roadmap, customer reference, or legal statement.
It is a personal body-of-knowledge document maintained by Pat Wendorf. It is based on anonymized field experience, technical discussions, customer patterns, and practical architecture lessons. Company names, project names, and identifiable details are intentionally avoided.
For official product documentation, use MongoDB and Voyage AI documentation directly.
Pull requests are welcome.
If you have experience building AI applications with MongoDB, Atlas Search, Atlas Vector Search, Voyage AI, or related architectures, feel free to contribute improvements, examples, clarifications, corrections, or additional patterns.
Good contributions might include:
- new anonymized use-case patterns
- clearer explanations of RAG, hybrid search, agent memory, or natural language query
- product capability corrections
- better caveats and limitations
- architecture diagrams or textual architecture patterns
- evaluation methodology improvements
- links to public MongoDB/Voyage resources
- typo fixes and readability improvements
Please keep contributions:
- anonymized
- non-confidential
- technically accurate
- respectful of customer privacy
- clear about what is experience/opinion vs. official product behavior
Do not include private customer names, internal project names, unreleased roadmap details, confidential metrics, private screenshots, or anything that would identify a specific customer or engagement.
The artifact this repository exists to publish is:
Start there.
This repository is released under the MIT License. See LICENSE.
Copyright (c) 2026 Pat Wendorf