August 2026 Summaries

Lovable can use Aiven’s Model Context Protocol (MCP) connector to provision and manage an Aiven for PostgreSQL database, create schemas and sample data, and deploy an application to Aiven Runtime while keeping database credentials hidden from the AI agent. After an Aiven organization administrator enables MCP access, Lovable connects through OAuth 2.0 and can access only the Aiven tools and permissions allowed by the user, with options to restrict its scope or make it read-only. In the demonstrated task-management application, the agent asked the user to select the Aiven project, PostgreSQL plan, cloud provider, and region before provisioning the service, then created and verified the application’s data layer through MCP calls. The deployed application requires its own database access, initially supplied through a Lovable secret but preferably through an Aiven service integration that injects and rotates a DATABASE_URL automatically. The approach places the database and application in the user’s Aiven account and chosen region, provides managed operational features such as backups and monitoring, and can later support services including Kafka, OpenSearch, and Valkey, although creating the Git repository required for Runtime deployment remains a manual Lovable action.

Aiven has introduced a new Apache Kafka pricing model for new customers on AWS, GCP, and Azure, while existing customers retain their current plans, prices, and contracts. The model prices services according to streaming-oriented measures—throughput, data retention, and topic type—using separate charges for cluster compute, object-storage capacity, and metered network ingress and egress, with diskless topics receiving lower network rates. Aiven says the change replaces machine-based plans that required users to estimate CPU, memory, and disk needs with more transparent pricing tied directly to measurable workload characteristics, such as MB/s and retention days. The company presents an example in which a workload formerly requiring an oversized hardware plan would cost less under the new structure, while costs scale gradually as traffic or retention increases. Network rates are flat and visible through a pricing calculator and service dashboards, and diskless topics, which shift replication to object storage, are now a standard capability that can be combined with classic topics on higher-throughput plans.

Dima Kan joined Aiven as product director for search after initially being considered for a Head of AI role, drawn by the company’s fast-moving, startup-like culture after working at larger organizations including TomTom and AlphaSense. Within his first week, he helped resolve a search-related issue for a high-stakes client deal, reinforcing his impression that Aiven enables rapid action and direct collaboration. Although he observed substantial leadership turnover shortly after joining, he focused on building relationships with his own team and adapting to the company’s ongoing change. Kan also introduced an engineer-to-engineer approach that connects Aiven engineers directly with client engineers to help unblock deals and better serve developer users. He believes Aiven is best suited to people who seek challenge and change rather than stability, and he has also discovered an unexpected aptitude for marketing through his independently successful podcast about vector databases and search technology.

Aiven has extended the open-source OpenSearch Migration Assistant beyond its previous AWS-only scope by contributing GCP support and private networking capabilities through five Terraform-based pull requests. The Kubernetes-native tool migrates Elasticsearch, OpenSearch, and Apache Solr workloads to OpenSearch by handling metadata and index mapping migration, live traffic capture and replay, and target-cluster validation. GCP deployments can now run on Google Kubernetes Engine, while private networking enables organizations in regulated sectors such as finance, healthcare, and government to keep migration traffic off the public internet. The contribution creates a supported migration route for GCP users moving from self-managed or hosted Elasticsearch and OpenSearch environments to Aiven for OpenSearch, a managed service available across AWS, GCP, and Azure, and Aiven plans to extend the Terraform approach to AWS support as well.

Apache Kafka Streams is a Java library for building scalable, fault-tolerant stream-processing applications that read data from Kafka topics, transform it through operations such as filtering, enrichment, or anomaly detection, and optionally write results to new topics. An Aiven Labs example repository provides a documented, vendor-neutral implementation using Java 25, Gradle, Avro messages in Confluent Wire Format, and both GenericAvroSerde and SpecificAvroSerde approaches, illustrating the trade-off between runtime schema flexibility and compile-time type safety. Its four ready-to-run applications demonstrate logging messages, copying messages unchanged, and filtering logistics data for delivered items while mapping it to a new output schema. The repository includes tests, a Dockerfile, Compose configurations, and instructions for local execution or deployment through Aiven Apps, where users can connect a Kafka service and Karapace schema registry, generate sample logistics data, configure application and topic variables, monitor build and runtime logs, and inspect processed Avro messages.

Contributing to large open-source projects like Apache Kafka presents challenges due to its vast and complex codebase, which spans multiple parts, languages, and versions. AI coding assistants can be valuable tools, but their effectiveness depends on the implementation of "guardrails"—explicit rules that guide their actions to ensure safe and compliant contributions. These guardrails include understanding when a Kafka Improvement Proposal (KIP) is necessary before making changes, recognizing which languages and versions are used across different modules, and ensuring that code cleanups do not unintentionally alter existing behaviors. While AI can assist with tasks like drafting and checking consistency, the human contributor must possess a solid understanding of Kafka's core concepts and maintain oversight to differentiate between sound AI suggestions and misleading ones. This approach not only aids in navigating Kafka's vast landscape but also helps contributors deepen their understanding of the project's nuances and processes.

A food enthusiast, in search of an unforgettable anniversary dinner, decided to leverage Aiven for PostgreSQL and AI tools to create a personalized database of local restaurants, inspired by the MICHELIN Guide and a colleague's blog post. By storing the data in PostgreSQL and using postGIS for spatial queries, they were able to filter restaurants by distance and other attributes. The process involved using Aiven MCP and Claude Code for data importation, with AI assisting in schema development and upload scripts. Despite some challenges with older tool versions, prompt hygiene helped mitigate issues. The setup included installing Aiven MCP for Claude Code and updating preferences in the claude.md file to streamline the process. After successfully uploading and normalizing the data, the user explored it with PG Studio, verifying schema accuracy and running complex queries with postGIS to narrow down restaurant choices within 100 miles of Atlanta. The final step was creating a materialized view to facilitate future queries for dining recommendations, resulting in a reusable tool that cost under $5 and took less than an hour to build. The experience was shared as a guide for others to replicate, inviting them to explore and expand upon this project using Aiven for PostgreSQL.