March 2026 Summaries

As software engineering evolves into agentic engineering, the integration of AI tools in development is becoming more prevalent, with 84% of developers using or planning to use such tools according to the Stack Overflow Developer Survey 2025. In response, MongoDB has introduced the MongoDB MCP Server and MongoDB Agent Skills to enhance coding agents' capabilities in generating more reliable code throughout the development lifecycle. These skills provide structured instructions and best practices for tasks like schema design, performance optimization, and advanced capabilities such as AI retrieval. By launching plugins for platforms like Claude Code, Cursor, Gemini CLI, and VS Code, MongoDB combines its MCP Server and Agent Skills into a comprehensive package, aiming to mitigate common coding pitfalls and enhance the quality of agent-generated code. The new tools ensure agents operate with necessary permissions and provide expert guidance to align with MongoDB's best practices, reducing architectural risks and accelerating implementation. As organizations adapt to agentic engineering, these tools facilitate faster and more reliable application development, embedding MongoDB's best practices into workflows while allowing for customization to enforce internal standards.

MongoDB has launched an Asia Pacific Strategic Partner Program aimed at transitioning the region from legacy infrastructure to modern, AI-driven applications by collaborating with select partners that possess deep architectural expertise and local knowledge. The initiative, which includes partners like Sieger, SoftwareOne, and ICS Compute, seeks to expand MongoDB's APAC partner team by 50% and quadruple strategic partnerships by the end of the fiscal year. The program provides partners with resources such as specialized certifications, localized expertise, modernization tools, and joint marketing strategies to facilitate AI adoption and modernization. The partnership has already shown results, with MongoDB aiding companies in China, India, and Indonesia to overcome legacy system challenges and establish scalable, AI-ready data architectures. MongoDB's strategy focuses on reducing delivery risks, shortening time-to-market, and supporting the modernization of core applications, ultimately enabling enterprises to effectively leverage AI and drive global success.

Andrei Radulescu-Banu, the creator of DocRouter.AI and SigAgent.AI, details the advantages of using MongoDB for AI applications that handle document- and log-heavy data. He explains that MongoDB is preferred over Postgres for applications requiring rapid schema evolution, horizontal scaling, and handling JSON-heavy AI workloads. The document-centric nature of MongoDB allows for quick iterations and schema flexibility, vital for AI product development, while also supporting vector search capabilities. The implementation involves disciplined use of migrations to ensure a predictable document structure, indexing for performance optimization, and a setup for handling vector searches. Andrei shares insights into a shared backend setup for DocRouter.AI, which routes documents and extracts structured data using LLMs, and SigAgent.AI, a monitoring agent, emphasizing the seamless integration and development efficiency achieved through MongoDB. He discusses a comprehensive approach to managing indexing, vector search, and reconciliation processes within knowledge bases, highlighting the use of a consistent data layer across different environments, including local, production, and self-hosted setups, to maintain operational simplicity and development speed.

Zomato, the world's second-largest food delivery company, has leveraged MongoDB Atlas to enhance its customer experience and operational efficiency through an AI-native platform called Nugget. Initially developed for internal use, Nugget automates high-volume customer support using real-time context storage, scalable data pipelines, and intelligent workflows, reducing support costs by 55% and increasing customer satisfaction. The platform, which was launched as an enterprise offering in 2025, uses MongoDB Atlas's flexible schema and high-performance capabilities to manage complex data and ensure compliance with security standards. By January 2026, Nugget had expanded its reach across industries like financial services, hospitality, and healthcare, handling over 15 million conversations monthly and improving query resolution rates by 85%. MongoDB Atlas's multi-tenant architecture allows Nugget to maintain data isolation while operating on dedicated database instances tailored to each customer's compliance needs, and Zomato continues to enhance Nugget with additional AI-driven features to further streamline customer support.

Voyage AI by MongoDB has focused on enhancing the efficiency of embedding models by introducing a mixture-of-experts (MoE) architecture in the voyage-4-large series, surpassing the limits of traditional dense models. Unlike dense embedding models where every token engages all parameters, MoE models use sparse layers with routers that direct tokens to specific expert networks, significantly reducing active parameters and computational demands without sacrificing retrieval accuracy. This approach allows the decoupling of computational cost from knowledge capacity, enabling the model to maintain high intelligence with reduced operational expenses. Key design choices, such as token dropping and router parameter management, were explored to optimize training throughput and model merging, respectively. The result is a 75% reduction in active parameters per token, offering the performance of a large model with the efficiency of a smaller one, demonstrating that the MoE architecture achieves comparable retrieval accuracy to dense models while drastically lowering inference costs and latency.

MongoDB has launched an enhanced data browsing experience within the MongoDB for Visual Studio Code extension, aiming to streamline developer workflows by integrating data management directly into the coding environment. As developers handle multiple tools daily, this new feature set seeks to reduce cognitive load by allowing users to browse, query, and edit data without leaving VS Code. The extension, now with nearly 3 million downloads, offers key improvements like a paginated tree view with prescriptive titles for easier data model comprehension, powerful action menus for seamless document management, and native editing with default Shell syntax to enhance compatibility with application code. These updates are designed to minimize disruptions to developers' flow states, enabling more efficient application development by reducing the need to switch between different tools. The new features are available immediately, with documentation provided to assist users in navigating the changes.

Modern digital lending platforms are revolutionizing the traditional lending process by integrating origination, decisioning, funding, and servicing into a seamless, intelligent workflow that meets the demands of today’s customers for instant decisions and transparency. As legacy systems are replaced by data-driven architectures, the lending lifecycle becomes a real-time, adaptive process driven by automation, AI, and integrated services, eliminating the manual, paper-heavy methods of the past. Agentic AI systems are pivotal in this transformation, actively managing tasks across the loan lifecycle, learning from new data, and adapting to changing conditions to optimize decisions, personalize borrower experiences, and enhance risk management. MongoDB supports this evolution by providing a flexible data platform that enables lenders to handle structured and unstructured data, ensuring scalability, resilience, and compliance. As digital lending enters a new phase characterized by proactive, self-optimizing systems, lenders that harness AI-powered platforms are positioned to lead the industry by anticipating customer needs and market shifts.

Global financial institutions have transitioned from the legacy SWIFT messaging standards to the ISO 20022 framework, which offers structured, machine-readable data for cross-border payments, significantly enhancing the clarity and detail of financial messages. However, the shift presents challenges, particularly when integrating ISO 20022's rich data into traditional relational databases, leading to increased ETL costs, performance issues, and compliance burdens. MongoDB has emerged as a key player in this transition, providing a flexible data platform that supports the complex, hierarchical structures of ISO 20022 messages without the constraints of conventional SQL databases. MongoDB Atlas offers solutions such as schema validation, real-time processing, and advanced search capabilities, enabling financial institutions to manage and query complex payment messages efficiently while maintaining data integrity and compliance. This transition underscores the need for modern data architectures that can adapt to evolving financial messaging standards, with MongoDB Atlas offering a comprehensive platform to support these requirements.

In enterprise environments, database outages can lead to significant business disruptions, such as failed transactions and inaccessible customer accounts, prompting teams to engage in labor-intensive diagnostic processes. MongoDB addresses this challenge by introducing a log integration that channels MongoDB Atlas system and audit logs directly into external observability and storage platforms like Datadog, Splunk, and Amazon S3. This integration allows database operators, DevOps professionals, and IT teams to export logs using OpenTelemetry, fostering seamless compatibility with various observability tools. The initiative aims to enhance real-time troubleshooting, unify telemetry by correlating logs with application and infrastructure data, and simplify compliance by automating audit log management. By enabling the rapid transmission and analysis of logs, MongoDB seeks to eliminate operational inefficiencies, reduce downtime, and maintain comprehensive visibility across an organization’s technology stack. This log integration is available for dedicated M10+ clusters and can be configured easily through the MongoDB Atlas interface, supporting scalable and effective observability strategies.

MongoDB Atlas Stream Processing now supports Apache Avro serialization integrated with the Confluent Schema Registry, which simplifies migrating streaming workloads by combining MongoDB's flexibility with the performance of binary serialization. This integration is particularly beneficial for use cases like real-time fraud detection, IoT sensor grid monitoring, and microservices synchronization, as it reduces the size of messages compared to JSON, thereby lowering bandwidth use and improving processing efficiency. Apache Avro, a compact and fast binary serialization format, uses an external schema for data interpretation, enhancing performance in high-throughput environments. The Confluent Schema Registry acts as a centralized source of truth for data structures, supporting decoupling, validation, and evolution of schemas to maintain reliable streaming pipelines. The integration allows MongoDB Atlas Stream Processing to automatically handle schema detection, caching, and deserialization of Avro-encoded data from Kafka topics, while also ensuring that outbound data conforms to predefined schemas. This capability eliminates the need for custom deserialization code, making it easier to build high-performance streaming pipelines without losing MongoDB's inherent flexibility.

Modelence is a comprehensive, AI-native development platform designed to streamline the creation and deployment of modern applications by offering all necessary components within a unified system. It facilitates rapid transitions from idea to production by integrating essential tools such as authentication, a MongoDB-powered database, live data systems, built-in monitoring, API endpoints, and deployment infrastructure. This platform eliminates the need for complex infrastructure management or additional services, allowing developers to focus on product logic. Modelence's App Builder, a central feature, enables the generation of complete applications from natural language prompts, embedding MongoDB best practices for a production-ready output. The platform supports seamless transitions from local development to cloud deployment, providing a consistent and simplified development environment. MongoDB, chosen for its flexibility and scalability, underpins these applications, ensuring efficient and hassle-free data management. Modelence offers practical examples of how its tools can be used to prototype and deploy applications efficiently, underscoring its value in creating full-stack, AI-enhanced applications.

MongoDB, a global technology company headquartered in New York, is celebrated not only for its innovative data solutions but also for fostering a thriving work culture and community, with a significant presence in Australia since 2012. The Sydney office, notable for its inspiring views and vibrant atmosphere, epitomizes MongoDB's commitment to employee well-being, growth, and inclusivity. Employees benefit from extensive wellness programs, flexible work models, and opportunities for professional development, creating an environment where ambition and collaboration flourish. Employees like Seemi and Satya highlight the supportive and empowering culture, which encourages them to take on complex challenges and contribute to meaningful projects like the Relational Migrator. This dedication to innovation and customer success is reflected in the company's culture of continuous learning and cross-functional collaboration, with initiatives that include speaker sessions, ERGs, and a focus on individual growth. MongoDB's unique organizational culture is marked by a commitment to diversity, creativity, and leadership development, making it a dynamic workplace where employees can make a global impact, as evidenced by successful product launches and long-term strategic partnerships.

The global semiconductor industry is witnessing rapid expansion, with sales projected to reach $975 billion by 2026 and $2 trillion by 2036, necessitating significant investments in semiconductor manufacturing equipment to accommodate technological advancements in AI, high-performance computing, and the automotive sector. However, legacy data infrastructures pose challenges in managing this growth, as fragmented systems and data silos disrupt critical workflows and increase costs. Companies are turning to AI and machine learning to harness about 40% of the value in manufacturing, but this requires a unified data backbone for efficient real-time detection and problem-solving. To address these challenges, the industry is adopting an agentic data layer architecture, exemplified by MongoDB Atlas, which combines document and vector databases to support varied data formats and enable AI agents to operate with real-time data access and autonomous actions. This approach consolidates data into a single platform, eliminating integration complexity and enhancing the ability to perform real-time anomaly detection, multimodal similarity searches, and agent-driven root cause analysis, ultimately transforming operational processes and improving manufacturing efficiencies.

Tokenization, a concept long discussed in financial services, involves transforming traditional assets into digital tokens to facilitate instant global trading. Although initial attempts struggled, recent developments, such as Robinhood's launch of tokenized U.S. stocks and ETFs in Europe, have demonstrated the commercial viability of tokenization for retail investors. This transition from experimental to foundational infrastructure offers numerous advantages, including increased liquidity, fractional ownership, faster and cheaper transactions, and enhanced transparency through blockchain technology. Tokenization also introduces challenges, such as regulatory uncertainty, cybersecurity risks, and the need for interoperability with legacy systems. MongoDB plays a critical role in supporting tokenization by providing a scalable, flexible, and real-time data architecture that integrates seamlessly with blockchain networks, enabling financial institutions to manage tokenized assets efficiently. As regulatory clarity improves and technical infrastructure advances, the financial sector is poised for significant growth in tokenized asset volumes, offering a competitive edge to firms that capitalize on these opportunities.

MongoDB has announced the general availability of standardized rate limiting for the Atlas Admin API v2, enhancing predictability and reliability for managing database clusters at scale. This update introduces a token bucket algorithm to standardize rate limits across API endpoints, providing clear operational boundaries while allowing flexibility for handling bursty traffic. The new system offers real-time visibility through standardized headers, such as RateLimit-Limit and RateLimit-Remaining, and supports resilient automation by allowing scripts to handle errors with intelligent retry logic. This change is designed to protect the Atlas ecosystem from surge loads while enabling developers and DevOps engineers to build smarter clients with predictable performance and reduced manual intervention. The update also includes guidance on handling "429 Too Many Requests" errors, suggesting strategies like exponential backoff with jitter for retrying requests. Users can audit their current usage through a new rateLimits endpoint and, if necessary, contact support to adjust rate limits for specific use cases.

Modern vehicles, acting as distributed computing systems, generate vast amounts of telemetry data, but traditional diagnostic and repair workflows still rely heavily on outdated methods like static documentation and keyword searches, often leading to incorrect repairs and increased costs. A significant issue is the fragmentation of diagnostic data across various formats, which simple search applications cannot effectively access. To address this, leading automotive organizations are adopting a unified architecture that integrates GraphRAG (the Relationship Engine) and Multimodal RAG (the Visual Engine) using MongoDB Atlas as a single operational data platform. This architecture allows for the ingestion, storage, and querying of technical documents, images, and system relationships, facilitating a more context-aware diagnostic approach that understands system relationships and visual content. The integration of GraphRAG and Multimodal RAG provides a comprehensive diagnostic intelligence platform that enhances technicians' ability to move from searching for information to effectively solving issues, thereby reducing repeat repairs and improving customer satisfaction.

The blog post by Arek Borucki, a Machine Learning Platform & Data Engineer at Hugging Face, introduces a tutorial for building a mood-based movie recommendation engine using various technologies, including the voyage-4-nano model, Hugging Face for model and dataset hosting, and MongoDB Atlas Vector Search. This innovative system allows users to search for movies based on their emotional state rather than traditional filters like genre or title. The tutorial details the use of AI embeddings to match user moods with movie plot descriptions, leveraging the voyage-4-nano model's embeddings, which are truncated to 1024 dimensions to balance semantic quality and storage efficiency. The tutorial also covers setting up a development environment with FastAPI and configuring MongoDB to manage connections and create a vector search index, while illustrating how Sentence Transformers can simplify working with embedding models. The post explains the process of downloading movie datasets, generating fresh embeddings, and implementing a search API that returns movies matching the user's mood, demonstrating how semantic search can capture nuances missed by traditional genre tags. The tutorial concludes with a comparison of embedding dimensions, highlighting how different dimensional settings impact retrieval quality and offering practical recommendations for balancing accuracy, storage, and latency.

MongoDB and DigitalOcean have expanded their strategic partnership to enhance the capabilities of developers and digital-native enterprises by offering a scalable, fully-managed, and cost-effective data platform solution. This collaboration integrates MongoDB's document model with DigitalOcean's managed cloud environment, helping customers rapidly deploy MongoDB-as-a-service clusters across DigitalOcean's extensive global data centers. Since the partnership's inception in 2020, over 9,000 customers have benefited from improved scalability and cost-efficiency. Notable examples include Jomashop, which improved its product data management and reduced infrastructure maintenance time by 70% using DigitalOcean's Managed MongoDB and Gradient AI Platform, and Picap, which reduced its monthly infrastructure costs significantly while increasing performance by leveraging DigitalOcean's Managed MongoDB and App Platform. The partnership aims to accelerate innovation in the AI era by continuing to enhance features such as Role-Based Access Control, Scalable Storage, and Autoscaling storage, thereby empowering the next wave of AI-native businesses and digital-native enterprises.

In a follow-up to a previous discussion on MongoDB's distributed transactions protocol, this text delves into the modular verification process used to ensure that the implementation of the WiredTiger storage engine aligns with its formal specification. By formalizing the interface between the distributed transactions protocol and the underlying storage engine, MongoDB developed a tool to automatically generate test cases to verify this conformance. Using a modified TLC model checker, the storage component's state space is explored to produce tens of thousands of test cases that verify the consistency of the WiredTiger implementation with the abstract model. This method not only provides a rigorous approach to verifying correctness across system layers but also opens up future possibilities for further exploration using techniques such as randomized path sampling. The article highlights the potential role of large language models (LLMs) in automating aspects of this verification process. More technical details and resources are available in the referenced VLDB ‘25 paper and associated GitHub repository.