1. What is daalu?

Intelligent IT operations — your whole estate in one workplace, paired with an AI assistant that does the work, not just talks about it.

daalu is an operations cockpit for teams who run real infrastructure. It puts your cloud accounts, your on-prem clusters, your network devices, and your observability stack into a single workplace — and pairs them with an AI assistant that can actually do work on your behalf, not just talk about it.

If you operate things — Kubernetes clusters, switches and routers, multi-cloud accounts, an alerting pipeline that pages you at 3 AM — Daalu is built for you.

The problem we’re solving

Modern operations teams live inside a tab gauntlet. A typical morning looks like this: AWS console for one cloud, GCP for another, Datadog for metrics, Grafana for the on-prem cluster, Slack for the alert that just fired, PagerDuty for the schedule, Nautobot for the network inventory, an Atlassian product for the ticket, a wiki for the runbook, and a terminal with eight kubectl contexts switched at the wrong moments.

Three things go wrong because of this fragmentation:

1. Investigation is slow. When something breaks, you don’t lack data — you lack the ability to correlate data across tools fast enough. By the time you’ve stitched together logs, metrics, and topology to form a hypothesis, the page has been open for an hour.

2. Action is even slower. Even after you know the fix, the distance between “I understand it” and “the fix is applied” is long: open a PR, get a review, wait for CI, hit the merge button, wait for the rollout, monitor it. For a small change that’s hours of friction; for an off-hours incident, it’s an eternity.

3. AI tools are mostly read-only chatbots. Today’s “AI for ops” tools can read your tickets, summarize your alerts, and write English about your dashboards. They cannot push config to your switch, restart a degraded service, or open a structured change proposal that your team can review.

Daalu attacks all three. The platform is integrated (all the sources in one place), it’s action-capable (the AI can mutate state, gated by human approval), and it’s opinionated (it ships with a way of working — alerts, incidents, change proposals, runbooks — so you’re not rebuilding the workflow yourself).

What Daalu actually does

Concretely, when you log in to daalu you get:

• A unified inventory. Cloud accounts, Kubernetes clusters, network devices, observability stacks, and CRMs all show up as first-class objects you can query, filter, and drill into.

• Alerts that explain themselves. When an alert fires, it arrives with a generated explanation, a suggested remediation, and a “page on-call” button — not a wall of labels. The explanation is grounded in the data Daalu can already see across your sources.

• An AI assistant with hands. The Daalu Assistant can query Prometheus, read pod logs, list cloud resources, propose a fix to a network device, or roll back a deployment. Every action that changes state is structured as a change proposal you approve before it executes. Every read is logged.

• A network source of truth. Daalu integrates with Nautobot (and a small set of similar systems). When the live state of a device drifts from the source of truth, Daalu opens a change proposal describing the drift — and once approved, it’s the same path that pushes the fix.

• Your own AI Factory. Connect an NVIDIA GPU and the AI Factory page lights up: every chat with the Assistant and every classifier call runs on your hardware first, with hosted LLMs as fallback. You pay for the GPU once and the AI calls are effectively free. The same page surfaces GPU health, utilization, and load-test results.

• A coding workspace. Once your GPU is online, the Workspace gives you a browser-based, AI-assisted coding environment backed by your own inference — useful for editing runbooks, automations, and infrastructure code without leaving daalu.

• One bill. Cloud spend, GPU usage, and daalu’s own subscription appear together under Usage & Pricing so you can see total cost per tenant rather than juggling vendor invoices.

The rest of this book is a tour of those capabilities. Chapter 2 explains the architecture at a high level so the rest makes sense. Chapters 4–9 take you from “I just signed up” to “my first cloud is connected, my GPU is serving inference, and I’m running a synthetic alert end-to-end.”

What Daalu is not

To set expectations:

• Daalu is not a replacement for Datadog or Grafana. It plugs into them — your Grafana dashboards stay where they are. Daalu’s job is to correlate across them, not to be a metrics store.

• Daalu is not a Slack-bot. There is a real web application with a real UI. Notifications go to Slack and PagerDuty, but the workflow lives in Daalu.

• Daalu is not a no-code platform. Operators are technical people who write YAML and read logs. The product is built for someone who already knows what kubectl get pods does — the AI is a force multiplier, not a substitute for the operator.

• Daalu is not a configuration management tool like Ansible or Salt. It doesn’t try to be the system that owns config. It is the system that coordinates and approves changes, then lets the owning system apply them.

How this handbook is organized

The book is laid out as if you were learning Daalu in order from nothing. Parts I–II are required reading for first-time users — they establish vocabulary and walk you through the first ten minutes. Part III explains the core concepts that every other chapter assumes. Part IV is a feature-by-feature tour you can read linearly or skip into. Parts V–VIII are reference: integrations, advanced workflows, pricing, and troubleshooting.

A glossary lives at the end. If a term is bolded the first time it appears (like tenant or change proposal), it’s defined in the glossary too.

Next: Chapter 2 — How it works