Home / Get started

Get started · about 4 minutes

Tell us how your estate is built.

A few questions about where your code lives, how it reaches your clusters and what may see your data. You see an honest fit as you answer, in the same Live, In development and Roadmap terms as the rest of this site. No credentials, no hostnames, and nothing gets connected.

We askYour tooling, platforms, rough counts and data rules.

We never askPasswords, tokens, kubeconfigs, hostnames or IP addresses.

ThenA reply, a call, and a read-only assessment of one repository.

Step 1 of 7 · about 1 minute

Where your code lives

The Git server that holds your Kubernetes manifests, Helm values or Flux config. That is where our draft merge requests would land.

Where are your deployment repositories hosted?

If you use several Git servers, pick the one that holds most of your cluster configuration.

GitLab, self-managed LiveOn your own servers, or GitLab Dedicated. GitLab.com LiveGitLab's SaaS. GitHub.com In developmentIncluding Enterprise Cloud. GitHub Enterprise Server In developmentOn your own servers. Forgejo or Gitea In developmentSelf-hosted. Bitbucket RoadmapCloud or Data Center. Azure DevOps Repos Roadmap Something else Ask us

Choose one option to continue.

Which GitLab tier?

It changes how our bot is let into your clusters. On Free, the GitLab agent acts as itself; on Premium and Ultimate it can act per user. Not sure? Pick that and we check on the call.

Free Premium Ultimate GitLab Dedicated Not sure

Choose one option to continue.

Can our bot reach your Git server from the internet?

Our bot calls your Git server's API from our side. If you only allow listed IP addresses, we have to run it from a fixed address.

Yes, no IP restrictions Live Only from allow-listed IP addresses Roadmap Only through a VPN or private network Ask us Not sure Ask us

Choose one option to continue.

How many repositories hold your cluster configuration?

Count the repositories Flux reads from. Leave out application source code.

1 2 to 5 6 to 20 More than 20

Choose one option to continue.

How are they organised? optional

Roughly is fine. The assessment maps it exactly.

One repository per cluster or environment Live One repository for several clusters In development A Flux hub cluster drives spoke clusters In development Not sure Ask us

Choose one option to continue.

Can we have our own bot account on your Git server?

What we prefer: we create a dedicated account for you, you invite it to the repositories, and no password or token changes hands. On paid GitLab tiers it may use a seat.

Yes, you can have your own account Live We create accounts ourselves and issue the tokens Live I need to check Ask us

Choose one option to continue.

Step 2 of 7 · about 30 seconds

How you deliver

What applies the configuration in git to your clusters.

What applies the configuration in git to your clusters?

Pick the main one.

Flux CD 2.x Live Flux v1 and v2 side by side Live Flux v1 only Live Argo CD RoadmapIncluding OpenShift GitOps. Both Flux and Argo CD Roadmap CI pipelines or people run helm and kubectl Ask usNo GitOps controller. Another controller Ask usFleet, Kluctl and others.

Choose one option to continue.

Is any of this also in your repositories? optional

For later. It does not change the first weeks.

Terraform or OpenTofu Roadmap Ansible Roadmap Pulumi Ask us

Choose at least one option to continue.

Step 3 of 7 · about 1 minute

Your clusters

We detect the exact flavour ourselves later. This tells us what to prepare and how our bot could read them.

Which Kubernetes platforms do you run?

Select all that apply.

Google GKE Standard Live Google GKE Autopilot In development Amazon EKS In development Azure AKS In development Red Hat OpenShift In developmentOn vSphere, bare metal, ARO or ROSA. Rancher RKE2 In development k3s In development SUSE Harvester Roadmap Upstream Kubernetes In developmentkubeadm, Talos, bare metal. Another managed service In developmentDigitalOcean, Scaleway, OVHcloud… Not sure Ask us

Choose at least one option to continue.

How many clusters should the assistant look at?

Production and non-production both count. You can start with one.

1 2 or 3 4 to 10 More than 10

Choose one option to continue.

How could our bot read your clusters?

Read access, listed verb by verb in a role file that sits in your own repository. With the GitLab agent your cluster dials out to GitLab: you open no inbound port and we hold no kubeconfig.

Through a dedicated GitLab agent LiveYou help us install it with one MR. Through the GitLab agent we already run RoadmapFor example the one CI/CD deploys with. Directly, from a fixed IP address RoadmapOur API server accepts allow-listed IPs. Only from inside our network Ask usVPN or bastion. No cluster access for now Ask usStart from git only. Not sure Ask us

Choose one option to continue.

Step 4 of 7 · optional · 30 seconds

Metrics and alerts

You can skip this. Metrics come later in onboarding, and nothing in the first weeks depends on them.

Where do your cluster metrics live? optional

Right-sizing and monitoring proposals need about 30 days of usage data.

Prometheus Live Mimir, Thanos or VictoriaMetrics Live Grafana Cloud Live Datadog, New Relic or Elastic Ask us Nothing yet Ask us

Choose at least one option to continue.

Could your clusters send a copy of their metrics to a private tenant we host? optional

One Grafana Alloy remote-write with a client certificate, one tenant per client. That is how right-sizing proposals get real numbers. We set it up with you later.

Yes We'd prefer you read from our backend No Not sure

Choose one option to continue.

Where do alerts go today? optional

For later. Reacting to alerts is switched on after go-live.

PagerDuty Live Alertmanager Live Opsgenie or Grafana OnCall Ask us Slack or Teams only Ask us Nowhere yet

Choose at least one option to continue.

Step 5 of 7 · about 1 minute

Data and AI

The read-only assessment uses no AI model at all. These answers matter from the first dry run on.

Where may an AI model read your configuration?

Secrets are always scrubbed before a model sees anything. This question is about everything else: names, hostnames, IP addresses.

A cloud model is fine Live Only with our names, hostnames and IPs masked In development Only EU-hosted providers Ask us Only on hardware we or you control In developmentNothing leaves. Not sure yet Ask us

Choose one option to continue.

Which AI providers may process it? optional

Every provider we would use is listed in writing as a sub-processor. You can narrow it further on the call.

Anthropic (Claude) Others on your sub-processor list Sent with the access letter. Decide on the call

Choose at least one option to continue.

Step 6 of 7 · about 1 minute

You and your goals

How to reach you, and what you would like the assistant to help with first.

What would you like to start with?

A read-only assessment of one repository A report on coverage, ownership and topology. Nothing is written anywhere. The managed assistant Starting with the same assessment. Monitoring and alerting improvements only

Choose one option to continue.

What should it help with first? optional

Pick up to three.

Keep charts, images and Flux current First weeks. Failing reconciliations and drift First weeks, one issue per cause. Requests and limits sized from real usage Needs metrics. Missing ServiceMonitors and dashboards Needs metrics. Manifest best practices After the first weeks. Getting off Flux v1, Tiller or hand-applied workloads Later. Security posture, Vault or GitLab reviews Arranged separately.

Choose at least one option to continue.

When would you like to start? optional

This month Within 3 months Just exploring

Choose one option to continue.

How do we reach you?

Your nameTell us what to call you.

Work emailWe reply only to this address. No newsletter.Enter an email address we can reply to.

Organisation optionalYour email domain usually tells us.

Your role optional

Anything else we should know? optionalPlease leave out passwords, tokens, hostnames and IP addresses. We ask for what we need, the right way.That looks like a credential. Remove it before sending, and if it is real, revoke it.

Leave this empty

I've read the privacy notice, and I may share this information about my organisation's setup.

We use these answers only to reply, to prepare the call and, if you go ahead, the read-only assessment. If we don't start working together, we delete them within 6 months. Before we read any repository, you get our data processing terms and the list of every sub-processor.

Please confirm to send your answers.

JavaScript is off. Send your answers by email instead: email us your answers →

Step 7 of 7 · check and send

Check your answers.

Nothing has been sent yet. Your fit below is provisional; we confirm it when we reply.

your-answers.yaml

These answers, your name and your email are all we receive. Nothing else is collected.

Received · reference

Thanks. Here is what happens next.

We read your answers, confirm the fit below and propose a call. Quote your reference if you write to us.

within 3 business days

Reply

We confirm the fit verdict and propose a call.

call · 45–60 min

Call and access letter

Your setup, your data rules, who reviews MRs. The same day: a one-page letter listing every permission we ask for, and our sub-processors.

target: 2 business days after read access

Read-only assessment

You invite our bot to one repository. The report comes from a process with no write path, no AI model and no tokens.

target: 2–5 business days, mostly your side

Onboarding MR

Grant roles, register a dedicated GitLab agent, merge one draft MR per deploy repository. Then a dry run that writes nothing.

a day after the dry-run sign-off

First draft MRs

At most 5 per agent per repository per run, labelled gitops-ai. Label one gitops-ai::pause and the bot stops at its next check.

What we will never ask

PWD
Passwords or tokens by email or a form.If someone asks you to email a token, it isn't us.
KCF
A kubeconfig.Cluster access goes through an agent you install and a role file you review, verb by verb.
OWN
Maintainer or Owner roles, or merge rights.Our bot opens drafts. Your team merges.
LOG
A login anywhere of ours.Everything arrives as merge requests and issues in your own Git server.

What was sent

Email them instead