{"id":109687,"date":"2026-01-08T22:33:56","date_gmt":"2026-01-08T22:33:56","guid":{"rendered":"https:\/\/mailinvest.blog\/index.php\/2026\/01\/08\/infrastructure-as-intent-oreilly\/"},"modified":"2026-01-08T22:35:15","modified_gmt":"2026-01-08T22:35:15","slug":"infrastructure-as-intent-oreilly","status":"publish","type":"post","link":"https:\/\/mailinvest.blog\/index.php\/2026\/01\/08\/infrastructure-as-intent-oreilly\/","title":{"rendered":"Infrastructure as Intent \u2013 O\u2019Reilly"},"content":{"rendered":"

<\/a>\n
<\/p>\n

\n

There\u2019s an open secret on this planet of DevOps: No person trusts the CMDB. The Configuration Administration Database (CMDB) is meant to be the\u00a0\u201csupply of reality\u201d\u2014the central map of each server, service, and utility in your enterprise. In concept, it\u2019s the muse for safety audits, price evaluation, and incident response. In observe, it\u2019s a piece of fiction. The second you populate a CMDB, it begins to rot. Engineers deploy a brand new microservice however overlook to register it. An autoscaling group spins up 20 new nodes, however the database solely information the unique three.\u00a0.\u00a0.\u00a0<\/p>\n

We name this\u00a0configuration drift<\/em>,\u00a0and for many years, our trade\u2019s answer has been to throw extra scripts on the downside. We write large, brittle ETL (Extract-Remodel-Load) pipelines that try to scrape the world and shove it right into a relational database. It by no means works. The \u201cworld\u201d\u2014particularly the trendy cloud native world\u2014strikes too quick.<\/p>\n

We realized we couldn\u2019t resolve this downside by writing higher scripts. We needed to change the elemental structure of how we sync knowledge. We stopped making an attempt to\u00a0boil the ocean\u00a0and repair your entire enterprise directly. As a substitute, we centered on one notoriously troublesome atmosphere:\u00a0Kubernetes<\/strong>. If we may construct an autonomous agent able to reasoning in regards to the complicated, ephemeral state of a Kubernetes cluster, we may show a sample that works in all places else. This text explores how we used the newly open-sourced\u00a0Codex CLI <\/strong>and theMannequin Context Protocol (MCP)<\/strong>\u00a0to construct that agent. Within the course of, we moved from passive code technology to lively infrastructure operation, remodeling the\u00a0\u201cstale CMDB\u201d\u00a0downside from a knowledge entry activity right into a logic puzzle.<\/p>\n

The Shift: From Code Technology to Infrastructure Operation with Codex CLI and MCP<\/h2>\n

The explanation most CMDB initiatives fail is ambition. They attempt to monitor each swap port, digital machine, and SaaS license concurrently. The result’s a knowledge swamp\u2014an excessive amount of noise, not sufficient sign. We took a distinct strategy. We drew a small circle round a selected area:\u00a0Kubernetes workloads<\/strong>. Kubernetes is the proper testing floor for AI brokers as a result of it\u2019s high-velocity and declarative. Issues change consistently. Pods die; deployments roll over; companies change selectors. A static script struggles to tell apart between a\u00a0CrashLoopBackOff<\/em>\u00a0(a short lived error state) and a purposeful scale-down. We hypothesized that a big language mannequin (LLM), appearing as an operator, may perceive this nuance. It wouldn\u2019t simply copy knowledge; it will\u00a0interpret<\/em>\u00a0it.<\/p>\n

The\u00a0Codex CLI<\/a>\u00a0turned this speculation right into a tangible structure by enabling a shift from \u201ccode technology\u201d to \u201cinfrastructure operation.\u201d As a substitute of treating the LLM as a junior programmer that writes scripts for people to assessment and run, Codex empowers the mannequin to execute code itself. We offer it with instruments\u2014executable capabilities that act as its fingers and eyes\u2014by way of the\u00a0Model Context Protocol<\/a>. MCP defines a transparent interface between the AI mannequin and the skin world, permitting us to reveal high-level capabilities like\u00a0cmdb_stage_transaction<\/em>\u00a0with out educating the mannequin the complicated inside API of our CMDB. The mannequin learns to make use of the software, not the underlying API.<\/p>\n

The structure of company<\/h3>\n

Our system, which we name k8s-agent, consists of three distinct layers. This isn\u2019t a single script operating high to backside; it\u2019s a cognitive structure.<\/p>\n

The cognitive layer (Codex + contextual directions)<\/strong>:\u00a0That is the Codex CLI operating a selected system immediate. We don\u2019t fine-tune the mannequin weights. Infrastructure strikes too quick for fine-tuning: A mannequin educated on Kubernetes v1.25 can be hallucinating by v1.30. As a substitute, we use\u00a0context engineering<\/a>\u2014the artwork of designing the atmosphere through which the AI operates. This entails software design (creating atomic, deterministic capabilities), immediate structure (structuring the system immediate), and knowledge structure (deciding what data to cover or expose). We feed the mannequin a persistent\u00a0context file<\/em>\u00a0(AGENTS.md) that defines its persona:\u00a0\u201cYou’re a meticulous infrastructure auditor. Your aim is to make sure the CMDB precisely displays the state of the Kubernetes cluster. You will need to prioritize security: Don’t delete information until you’ve gotten constructive affirmation; they’re orphans.\u201d<\/p>\n

The software layer<\/strong>:\u00a0Utilizing MCP, we expose deterministic Python capabilities to the agent.<\/p>\n