June 2026 Summaries

Claude Routines are automation tools designed to run on Claude's infrastructure, allowing tasks to be executed without human intervention. These routines can be triggered by schedules, API calls, or GitHub events, providing a persistent automation mechanism that was traditionally achieved using cron jobs and servers. However, the decentralized nature of configuring network access per engineer and project poses significant security risks, as any individual can expand an agent's access, creating a broad attack surface without centralized oversight. The article argues for the necessity of centralizing network governance to mitigate these risks, suggesting that automation should be managed within a governed runtime environment where a platform team enforces network rules through role-based access control (RBAC), domain allowlists, and auditing. By doing so, the platform team can ensure that automation remains secure and within controlled boundaries, preventing unauthorized access and potential vulnerabilities.

An MCP server, or Model Context Protocol server, functions as a standardized interface enabling AI agents like Claude Code and Cursor to interact with real infrastructure, such as deploying applications and provisioning databases, through a single governed API. This setup resolves the challenge of AI agents being able to generate code but struggling to operate the fragmented infrastructure stack, which includes systems like CI/CD and Kubernetes, each with its own interface. The core advantage of an MCP server lies in its ability to provide a consistent protocol for agents to execute infrastructure operations, while ensuring governance through features like role-based access control (RBAC), budget limits, and audit trails to prevent security incidents and unexpected costs. By integrating an MCP server, AI agents can safely and efficiently manage infrastructure tasks, while maintaining control and visibility for platform engineering teams.

In 2026, Base44 and Lovable emerge as competitive AI app builders that allow users to create web applications by simply describing them in plain English, with each platform catering to different needs and preferences. Lovable, founded in Stockholm in 2023, is the market leader with approximately 8 million users, $200M in annual recurring revenue, and a $6.6B valuation, known for producing highly polished React + Supabase apps and hosting a large ecosystem and community. Meanwhile, Base44, acquired by Wix in 2025, offers a lean, all-in-one solution with built-in features like user authentication, databases, and autonomous "Superagents," appealing to users who prefer simplicity and predictable pricing. While Lovable boasts a more extensive template library and community, Base44 is favored for its ease of use and integration-free setup. Both platforms initially host apps on their shared cloud, making them suitable for prototyping, but when transitioning to production environments requiring compliance and private networking, users need to deploy their apps on their own infrastructure using platforms like Qovery.

In 2026, Salesforce's decision to put Heroku into maintenance mode and end new Enterprise sales has created a pivotal moment for companies to transition from Heroku to their own cloud infrastructure. The advent of AI-driven migration tools has significantly reduced the complexity and time required for such transitions, allowing migrations that once took months to be completed in mere hours. As many engineering teams pivot towards AI-centric work involving machine learning services and autonomous agents, Heroku's limited capabilities and lack of future development make it an unsuitable platform. By moving to their own cloud accounts, organizations can maintain developer-friendly environments while gaining access to advanced infrastructure like Kubernetes, SOC2-grade isolation, and GPU workloads essential for modern AI applications. The current landscape, characterized by Heroku's stagnation and the rise of agentic infrastructure, presents an ideal opportunity for companies to make this transition before being locked into another year with a platform that no longer supports their growth ambitions.

In 2026, AI coding agent sandboxes are isolated environments where autonomous coding agents can write and execute code without affecting production systems, addressing security risks but often limiting verification capabilities. The landscape of AI coding agent sandboxes divides into agent-native sandboxes such as OpenAI Codex, Cursor, Claude Code, and GitHub Copilot, which prioritize safety by isolating environments, and infrastructure platforms like Qovery, which provide governed, full-stack environments for agents to deploy, test, and verify their code. While sandboxes effectively handle code generation, they typically fall short in enabling agents to deploy and conduct end-to-end tests in real environments, leaving the final verification to manual review. Qovery stands out by offering an agent-agnostic platform that allows AI agents to operate in a governed environment on a Kubernetes cluster, supporting deployment and verification with built-in governance measures like DLP filters and audit trails. This approach aims to bridge the gap between secure isolation and the need for comprehensive verification in real-world settings.

As Kubernetes, originally designed for stateless workloads, increasingly intersects with AI agents that require persistent memory and complex scheduling, the integration of AI in Kubernetes infrastructure is becoming crucial by 2026. This guide highlights two primary use cases: utilizing AI agents to manage Kubernetes (AIOps) and running AI agent workloads on Kubernetes. Tools like Qovery AI Copilot and Botkube facilitate cluster management by using AI to streamline operational tasks, while platforms like Sedai and K8sGPT optimize resources and diagnose issues. Meanwhile, for hosting AI workloads, solutions such as Qovery and KubeRay manage Kubernetes complexities, offering features like GPU scheduling and multi-service architecture support. Helm and ArgoCD provide a GitOps stack for declarative deployments, with Qovery offering an abstraction layer to enhance development efficiency. These tools aim to simplify Kubernetes operations, cater to the unique needs of AI workloads, and ensure secure and compliant deployments.

AI agents are revolutionizing Kubernetes cluster management by drastically reducing incident diagnosis time from up to 45 minutes to mere seconds and eliminating the manual burden of YAML authoring, shifting resource tuning from a static to a continuous process. They enable faster incident diagnosis by providing plain-English root-cause analysis through tools like K8sGPT, allowing engineers to focus on decision-making rather than problem identification. Continuous resource optimization is achieved as AI agents autonomously adjust configurations based on real-time telemetry, reducing cloud costs. Platforms like Qovery further enhance efficiency by allowing developers to deploy environments through natural language prompts, avoiding the need for YAML configuration. AI-driven scaling anticipates load demands, ensuring smooth handling of traffic spikes, while self-service operations empower non-platform engineers by reducing dependency on specialized knowledge. However, AI agents complement rather than replace essential infrastructure practices such as RBAC, capacity planning, and network topology decisions.

Qovery has evolved into an agentic platform as a response to market demand, driven by the challenges of integrating AI agents with existing infrastructure. As companies increasingly adopt AI, traditional tools like GitOps and Terraform, designed for human workflows, struggle to handle the volume and complexity of agentic operations. This transition has led to infrastructure breakdowns and highlighted the need for governance, security, and compliance, especially in regulated industries. Qovery addresses these issues by providing a centralized control plane that centralizes operations through a governed API, ensuring auditability and security while maintaining separation between the control and data planes. This architecture supports the safe integration of AI agents by offering scoped, authenticated access, and a complete audit trail, thereby fulfilling the requirements of compliance frameworks such as SOC 2, HIPAA, and GDPR. The company's platform allows both technical and non-technical users to leverage AI capabilities while maintaining control over infrastructure, positioning Qovery as an essential solution in the evolving landscape of AI-driven operations.

An agentic infrastructure platform represents a transformative approach designed specifically for AI agents, encompassing a unified API that integrates the entire infrastructure stack, including databases, networking, and CI/CD, to facilitate seamless operations. Unlike traditional platforms built for human interaction, where infrastructure operations are manual and fragmented, this new model prioritizes programmatic access and governance, thereby addressing the critical bottleneck caused by AI agents interacting with human-centric systems. The platform ensures robust governance with features like RBAC for non-human actors, budget controls, and comprehensive audit trails, which are essential for maintaining compliance and operational integrity, especially as agent-initiated operations surpass human capabilities. By providing an environment-native and agent-agnostic framework, it allows for full lifecycle management, enabling efficient scaling and cost control, thereby preventing issues like infrastructure sprawl and unanticipated expenses. As AI agents become more dominant in initiating infrastructure tasks, companies adopting agentic platforms will gain a significant advantage in AI-driven development, while those relying on outdated, human-oriented tools will face inefficiencies and increased operational risks.

The Qovery Console has undergone a significant redesign to improve navigation, scalability, and user experience. This revamp addresses previous challenges of a cluttered interface that became difficult to navigate as new features were added. By moving the navigation to the header, providing overview pages at every level, and implementing a modern frontend architecture with TanStack Router and React Suspense, the Console now offers a more intuitive and seamless experience. The redesign also simplified the token system, enabling features like dark mode, and shifted towards a domain-oriented structure to enhance maintainability. Feedback from users has been positive, highlighting the improved consistency and navigation, while the open-source nature of the Console allows for community engagement and continuous improvement.

At Devoxx France 2026, the focus was on the disciplined application of AI in software development, emphasizing key themes such as context engineering, change management, cost optimization, and sovereignty. The event highlighted the importance of harness engineering, where context files and governance become production artifacts, enabling AI to align with team knowledge and improve task performance. It was observed that AI-generated pull requests have lower acceptance rates compared to human-generated ones, indicating a need for robust review and deployment pipelines. Additionally, the conference underscored the significance of localized and secure AI agent deployment, as remote models pose security and cost challenges, especially with evolving regulations. Overall, the discipline that once seemed optional has now become essential in managing AI's impact on software engineering practices.

Integration of Secret Manager with Qovery offers a solution for managing production secrets by connecting AWS Secrets Manager, AWS SSM, or GCP Secret Manager directly to the platform, ensuring secrets are referenced rather than copied. This integration addresses compliance issues and security risks associated with duplicating secrets across systems by maintaining a single source of truth within the customer's cloud boundary. It simplifies the deployment workflow by allowing developers to bind secrets as environment variables without additional infrastructure management or changes to existing processes. The integration supports enterprise-grade governance, recording all bindings in an audit trail and applying consistent rotation policies and access controls, even when deployments are triggered by AI agents. The updated Qovery interface includes a reorganized environment variables panel and a new cluster Add-ons section to streamline management and enhance the user experience.

AI coding agents like OpenAI's Codex and GitHub Copilot are becoming increasingly competent at generating code based on specifications, yet they often fall short in the end-to-end software development process, particularly in deployment and verification. These agents typically operate in isolated environments, which, while enhancing security, limit their ability to interact with real-world infrastructures where most bugs and integration issues arise. This results in a broken development loop, as human developers must manually deploy and test the code in genuine environments. Qovery addresses this gap by enabling coding agents to operate within real infrastructure environments on Kubernetes clusters, allowing them to deploy, test, and verify their code autonomously while maintaining governance and security. By providing agents with real operational contexts, Qovery aims to streamline the development process, transforming the human role from hands-on verification to reviewing functional software through preview URLs. This advancement seeks to fully integrate coding agents into the software development lifecycle, bridging the gap between code generation and real-world implementation.

Antoine Promerova, a senior software engineer at Qovery, shares his experience of constructing a private cloud on bare-metal infrastructure to better understand the complexities of cloud platforms like Qovery. This hands-on project involved setting up an Incus cluster with Kubernetes (K3s), OVN networking, and modern ingress solutions, bypassing traditional NGINX controllers in favor of the Kubernetes Gateway API. Promerova encountered numerous challenges, such as network tunneling issues and CI/CD pipeline maintenance, highlighting the significant time and effort required to manage infrastructure manually. Despite the educational benefits of this endeavor, he underscores the advantage of using platforms like Qovery that automate complex infrastructure tasks, allowing developers to focus on software deployment and value delivery rather than infrastructure troubleshooting.

OpenAI's Codex has reached 5 million weekly users, with 20% being non-developers, whose usage is growing three times faster than that of developers. This growth supports the thesis that AI coding tools are transforming non-technical roles into software builders. OpenAI introduced six role-specific plugins for various departments, like finance and marketing, enhancing productivity by integrating directly with tools like Snowflake and Salesforce. However, the lack of an enterprise governance layer for these tools poses significant challenges for regulated industries, as OpenAI’s infrastructure currently lacks controls such as SSO, RBAC, and data residency assurances. The article argues for the need for a governed runtime environment, suggesting that platform engineering teams should build this infrastructure to provide the necessary governance while allowing employees to leverage AI tools effectively.