2025 Year-End Product Update: The Year AI Came to DevOps

What's new in Plural

We spent 2025 building AI features. The goal wasn't answering questions about Kubernetes but automating actual infrastructure work. Diagnosing failures, generating Terraform changes, validating cluster upgrades. Each feature required treating our GitOps platform as a structured knowledge base the AI could query and modify.

The result: platform teams can now diagnose issues in seconds instead of hours, generate infrastructure changes with natural language, and validate upgrades before they break production. Here's what we shipped.

2025 recap: AI that actually understands your stack

This year's biggest bet paid off. We didn't just add AI features. We made Plural AI-native. Every capability below shares the same foundation: semantic understanding of your entire infrastructure, strict permission boundaries, and GitOps-based review loops that keep humans in control.

AI insights: automatic inspection that finds the real problem

Your monitoring shows 66 of 67 services healthy. Users are screaming. Alerts fire everywhere. Which component is actually broken?

We built AI Insights to answer that question instantly. When something fails, Plural's AI Insight Engine automatically collects logs, events, and configuration data across your Kubernetes clusters. It analyzes the full context, not just the symptom, and generates a root cause summary with evidence, debugging steps, and a PR to fix the issue.

The system handles everything from ImagePullBackOff errors to Terraform state corruption. It parses verbose Terraform logs, identifies invalid parameters, and creates targeted fixes. For complex issues, a full chat interface lets you dig deeper with the AI while it maintains context about your specific stack.

Semantic search: every resource, instantly findable

Finding that one ConfigMap across 40 repositories used to mean grep, guesswork, and asking the one engineer who remembers where it lives.

We introduced AI auto-vectorization that semantically indexes your entire infrastructure. Every Terraform state graph, Kubernetes manifest, Helm chart, and service definition gets vectorized along with its relationships. The AI develops contextual understanding of how your components connect, not just what they're named.

Ask "show me everything related to our payment service" and get results spanning multiple repos, clusters, and Terraform modules. The search is permission-aware: you only see what your RBAC allows, even when querying live clusters.

Coding agents: natural language to pull request

This is where things get interesting. Our infrastructure coding agents don't just answer questions. They make changes.

Tell Plural to "double the size of my production database" and watch what happens. The agent traverses your service catalog, inspects your cluster state, finds the relevant Terraform or Kubernetes manifests, and generates the necessary changes. It creates a PR, reviews the Terraform plan output, and self-corrects if something looks wrong. You review, approve, and merge.

Three modes work together:

• Search mode finds resources across repositories and queries live Kubernetes clusters
• Provisioning mode guides you through service catalogs and PR automation workflows
• Manifest generation creates accurate YAML for any operator by querying the Kubernetes API discovery endpoint

The secret sauce is our GitOps-based RAG engine. Because Plural already manages your deployments, the AI has complete context about your infrastructure, not just code snippets, but live state, deployment history, and relationships between components.

Infra Research: architecture diagrams that stay current

The architecture of a complex system shouldn't exist only in the minds of three senior engineers. That bottleneck slows incident response, makes onboarding painful, and contributes directly to burnout.

Plural Infra Research is an AI agent that automatically diagrams and analyzes your infrastructure from GitOps repositories. Give it a prompt like "show me the architecture of the Grafana deployment" and it iteratively builds a complete picture of your stack.

The agent indexes your vectorized infrastructure data, queries it iteratively to build a knowledge graph, and synthesizes everything into a Mermaid diagram with plain-English analysis. It identifies components, connections, dependencies, and even potential misconfigurations.

What makes this different: Infra Research connects your application layer (Kubernetes) with your infrastructure layer (Terraform). It sees your VPC, RDS instances, OIDC providers, not just deployments and services. Published diagrams become trusted artifacts for team sharing and continuously improve the AI's understanding.

Sentinels: infrastructure testing that scales

Manual runbooks don't scale. Checklists get skipped. Post-deployment "smoke tests" catch problems too late.

Plural Sentinels brings automated, deep integration testing to your delivery pipeline. This hybrid AI/non-AI system packages infrastructure validation into auditable, repeatable tests that run before changes hit production.

Sentinels validate that your infrastructure actually works the way you expect, not just that manifests apply successfully. They catch the subtle issues that slip through linting and static analysis: misconfigured network policies, broken service mesh routing, permission gaps between environments.

AI-powered stack approval: guardrails that understand context

Manual approvals for every infrastructure change create bottlenecks. Approving everything automatically creates risk. We needed something smarter.

AI-powered stack approval analyzes proposed changes against your defined rules and organizational context. Low-risk changes that match established patterns get auto-approved. Unusual changes, sensitive resources, or production modifications route to the right reviewers with AI-generated summaries of what's changing and why it matters.

The system learns from your approval patterns over time, getting better at distinguishing routine updates from changes that need human judgment.

2025 recap: everything else we shipped

Upgrade assistant: deterministic recommendations, not guesswork

Kubernetes upgrades remain one of the highest-risk operations in infrastructure. Our enhanced upgrade assistant now provides deterministic recommendations based on your actual cluster state.

The system scrapes cloud provider endpoints (especially EKS) for upgrade intelligence, tracks Helm chart compatibility across versions, and flags deprecated APIs before they break. It identifies add-on incompatibilities, suggests safe upgrade paths, and generates GitOps-ready remediations.

New this year: the assistant now recommends specific Helm chart version changes required for your target Kubernetes version. No more hunting through changelogs or discovering incompatibilities in production.

Plural Flows: your team's microservice command center

Platform engineers and application developers need different things. Flows gives developers a focused interface for managing the services they own.

A Flow is a logical container that groups related microservices and their development pipelines. Within a Flow, teams can monitor service health, track deployments, see related PRs, and troubleshoot issues, all within strict permission boundaries they don't have to think about.

Flow Chat provides AI-powered troubleshooting scoped to the Flow's services. It analyzes alerts and logs, cross-references recent PRs from GitHub or GitLab, identifies exact code changes causing issues, and pinpoints specific commits and lines of code. Developers get targeted fix recommendations without needing to understand the entire platform.

MCP server integration lets teams connect third-party tools for automation without building custom admin interfaces.

2026 outlook: what's coming next

Based on what we've learned shipping AI features this year, here's where we're headed.

Claude Code and API-driven coding agents

We're integrating Claude's coding capabilities via API, running agents directly on your Kubernetes infrastructure. This means more sophisticated code generation, better understanding of complex codebases, and tighter integration with your existing development workflows.

Automatic CVE remediation

When a critical vulnerability drops, you shouldn't need an all-hands fire drill. We're building API-driven CVE remediation that automatically identifies affected components, generates patches, and creates PRs, cutting response time from days to minutes.

Slack integration for Plural AI

Your team lives in Slack. Soon you'll trigger Plural AI directly from channels and threads. Ask questions about infrastructure, kick off investigations, or request changes without leaving your conversation.

Ticketing system integrations

Jira and Linear integrations will connect infrastructure work to your existing project management. AI agents can be triggered from tickets, and remediation PRs can automatically link back to the issues that spawned them.

Incident remediation agent

When PagerDuty fires, our incident remediation agent will automatically investigate alerts, correlate symptoms across your stack, and surface probable root causes with suggested fixes. The goal: resolve incidents faster by doing the initial investigation before a human even looks at the alert.

On-prem Kubernetes provisioner

For teams that can't use cloud-managed Kubernetes, we're building an on-prem provisioner using Cluster API. Declarative cluster lifecycle management for bare metal and private cloud environments, with the same GitOps workflows you use everywhere else.

REST API with full OpenAPI schema

A complete REST API with OpenAPI specification for everything in Plural. Build custom integrations, automate workflows, and extend the platform however you need.

Try it yourself

2025 proved that AI can do more than autocomplete YAML. With the right context and guardrails, it can fundamentally change how teams operate infrastructure.

If you're managing more than a handful of Kubernetes clusters and want to see what AI-native operations looks like, book a demo or start a free trial.

For technical deep-dives on anything covered here, check our documentation or drop into our Discord.