# MongoDB GenAI Showcase: Official RAG and Agent Example Repository > This official MongoDB collection of Jupyter notebooks, JavaScript and Python apps, and self-paced workshops covers retrieval-augmented generation, AI agents, and industry-specific use cases with Atlas as the database layer. It is organized for engineers who are just starting out with generative AI as well as those building production-scale agentic systems. **mongodb-developer/GenAI-Showcase** — MongoDB's Generative AI Showcase: an exhaustive collection of examples and sample applications covering Retrieval-Augmented Generation (RAG), AI agents, and industry-specific use cases. - Repository: https://github.com/mongodb-developer/GenAI-Showcase - Website: https://www.mongodb.com/cloud/atlas/register - Stars: 4,265 · Forks: 745 - Language: Jupyter Notebook - License: MIT - Published: 2026-09-10 · Updated: 2026-09-10 · Language: en - Canonical page: https://hysenlabs.com/projects/mongodb-developer-genai-showcase ## What the repository addresses and who it serves Building a retrieval-augmented generation pipeline or an AI agent requires an operational database, a vector store, and often a place to persist conversation memory. Engineers who want to understand how MongoDB Atlas fits into each of those roles have historically needed to assemble that picture from scattered blog posts and documentation pages. This repository puts working examples in one place. The collection is organized around three core topics: RAG, AI agents, and industry-specific use cases. The README describes Atlas as serving three distinct functions inside these patterns: as a vector database that stores embeddings for similarity search, as an operational database that stores the raw documents and metadata alongside those embeddings, and as a memory provider that holds agent conversation state across turns. The target audience spans beginners starting a generative AI project for the first time through to teams building production-scale agentic systems. ## Repository layout: four folders and what each contains The repository has four primary content directories. The notebooks folder holds Jupyter notebook examples covering RAG implementations, agentic application patterns, and evaluation approaches. The apps folder contains both JavaScript and Python applications and demos. The workshops folder offers self-paced hands-on labs designed for engineers who prefer a structured learning path over individual notebooks. The partners folder holds contributions from MongoDB's AI partner ecosystem. Beyond those four, the top level also contains a mcp/ directory, a misc/ directory, a tools/ directory, and a resources/ directory, alongside the AGENTS.md file. The presence of AGENTS.md indicates that AI-agent workflow tooling has been integrated into the development process. The ruff.toml at root applies Python linting rules across the Python portions of the repository. The repository has no GitHub releases; new examples land directly on the main branch. ## Getting started: Atlas cluster as the hard prerequisite Every example in the repository requires a live MongoDB Atlas cluster. The README describes three mandatory setup steps: register for a free MongoDB Atlas account at the Atlas registration page, create a new database cluster through the Atlas console, and obtain the connection string for that cluster. The repository itself carries no local-only execution path. There is no docker-compose file or self-hosted option described in the README. Engineers who prefer to avoid a cloud dependency before evaluating the examples will need to read the Atlas free-tier documentation separately. The README does not document a local MongoDB setup as a substitute for Atlas, which means the barrier to running even the simplest notebook includes a cloud account, a cluster, and a network-accessible connection string. ## Atlas as vector database, operational store, and agent memory The README explicitly names the three roles Atlas takes in the example patterns. As a vector database, Atlas stores the embeddings that RAG pipelines query during retrieval. As an operational database, it stores the source documents and their associated metadata in the same cluster alongside the vectors, so applications can retrieve both structured data and semantic search results from a single connection. As a memory provider, it persists the conversation history that agentic applications need to maintain context across multiple tool calls or model turns. This three-in-one positioning is the core argument of the collection. Rather than running a dedicated vector database separately from the application database and a separate key-value store for agent memory, the examples demonstrate a single Atlas cluster serving all three functions. Whether that consolidation fits a given architecture depends on the team's existing infrastructure and query volume patterns, which the examples do not address directly. ## What the example collection does not cover The repository contains no prebuilt installable package. There is no pip install or npm install command that pulls in the repository itself; each notebook or app must be cloned and run individually after Atlas is configured. The README does not document rollback procedures, environment teardown, or cost estimates for Atlas usage at scale. The partners folder introduces a quality-consistency question: external contributions follow their own structure and may not match the depth or testing level of the MongoDB-authored notebooks. The README does not describe a review standard for partner contributions. The collection also does not cover non-MongoDB vector stores, so engineers evaluating multiple vector databases will not find comparison notebooks here. ## How this compares to framework-level example collections LangChain maintains its own cookbook, a well-known open-source collection that covers RAG and agent patterns across multiple vector stores and model providers. The LangChain cookbook is framework-first: examples are organized around LangChain's abstraction layer and work with various underlying databases. This repository takes the opposite approach: it is database-first, with Atlas as the fixed component and a range of frameworks and model providers used across different examples. For a team that has already chosen MongoDB Atlas and wants to understand how to integrate a language model, the database-first organization here is a direct fit. For a team still evaluating which vector store to use, the LangChain cookbook or similar multi-provider collections give a wider comparative view. ## Maintenance and licensing terms The repository is published under the MIT license, which allows commercial and private use without restriction and does not require derivative works to adopt the same license. The last push was on 2026-09-04, and the repository accepts community contributions through the process described in the CONTRIBUTING.md file. There are no GitHub releases; content updates go directly to the main branch. The AGENTS.md file at the root suggests that the maintainers use AI agent tooling in their own workflow, which aligns with the subject matter of the collection. Support for issues raised while working through the examples is handled through the GitHub Issues tab, and MongoDB provides additional resources through its AI Learning Hub and the GenAI Community Forum linked in the README. ## Conclusion Engineers already running MongoDB Atlas for their application data will find this repository immediately useful: every notebook, app, and workshop targets Atlas directly rather than a generic store. Before committing, verify that you have an Atlas cluster configured with a working connection string, because the README states no example will run without one. Teams using a different vector database will find the examples harder to adapt, since Atlas-specific connection and indexing patterns appear throughout. ## FAQ ### What roles does MongoDB Atlas serve in these example notebooks? The README states that Atlas serves as a vector database for storing embeddings, as an operational database for storing raw documents and metadata alongside those embeddings, and as a memory provider for agent conversation state across turns. ### Can I run the GenAI example notebooks without a paid Atlas account? The README directs readers to register for a free MongoDB Atlas account and create a cluster before running any of the examples. The free-tier cluster is the expected starting point, though the README does not document a fully local alternative. ### Does the repository include examples for specific industries? The README describes industry-specific use cases as one of the three main topic areas alongside RAG and AI agents. The partners folder holds contributions from MongoDB's AI partner ecosystem, which adds further domain coverage beyond the MongoDB-authored examples. ## Sources - [Issues](https://github.com/mongodb-developer/GenAI-Showcase/issues) - [License: MIT](https://github.com/mongodb-developer/GenAI-Showcase/blob/main/LICENSE) - [mongodb-developer/GenAI-Showcase on GitHub](https://github.com/mongodb-developer/GenAI-Showcase) - [Project website](https://www.mongodb.com/cloud/atlas/register) - [README](https://github.com/mongodb-developer/GenAI-Showcase/blob/main/README.md) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/mongodb-developer-genai-showcase