Guided, hints are free
Tutorials
A step-by-step path, with notes and the editor open beside the terminal. To learn something new without getting stuck.
- Progress is saved
- Hints cost nothing
Private preview is open · sign-up by invitation
Real infrastructure.
Powered by your AI agents.
Klastro starts an isolated Kubernetes cluster or Linux machine in seconds. You do the work inside it, your result is graded automatically, and your AI agent can open labs and track your scores.
The technologies you actually operate during a session
Nobody learns to fix a cluster by ticking boxes
An engineer picking up Kubernetes has two unsatisfying options. Quiz platforms confirm they memorised the syntax of a manifest, never that they can bring a cluster back up. Demo labs lend them a terminal for twenty minutes and grade nothing: they leave without knowing whether they got it right.
The same gap exists in hiring. A technical interview ends on an impression, or on an algorithm exercise that says nothing about diagnosing a pod that refuses to start. Teams hiring cloud-native engineers know this, and work around it with homegrown tests that are slow to write and impossible to compare between candidates.
Those are two symptoms of one hole: there is no real, disposable infrastructure that a machine can grade. Klastro is that engine. It brings up a dedicated cluster in seconds, lets someone work in it without a safety net, then reads the cluster’s actual state to establish what is done and what is not.
A disposable cluster, graded on its real state. Learning and assessing are then just two ways of looking at the same session.
Quiz platforms test recall. Demo labs grade nothing.
Klastro sits at the intersection of the two: real isolated infrastructure, work that is graded automatically, and a single engine that serves both training and hiring.
| Real, isolated infrastructure | Quiz platforms | Demo labs | Klastro |
|---|---|---|---|
| Real, isolated infrastructure | – | ||
| Automatic, reproducible grading | – | ||
| Live supervision | – | – | |
| Per-session network isolation | – | ||
| One engine for training and hiring | – | – |
From scenario to result
Klastro handles the infrastructure. What is left is the work itself.
-
Pick a scenario
A catalogue of cloud-native scenarios, or your own. Each one defines its environment type, the tasks to complete, and the validation rules that grade them.
-
Open the environment
A dedicated cluster starts in seconds. Terminal, editor and notes open in the browser. Nothing to install locally.
-
Collect the result
Validation rules run against the real cluster state. Learners see their progress, hiring teams get a score they can defend.
Choose how you work
The same real infrastructure, framed around what you are trying to do.
Graded automatically
Challenges
A failure to fix, an objective to reach. Validation rules read the real cluster state and the score lands immediately.
- Visible score
- Attempt history
Open, no objective
Playgrounds
An empty cluster, yours. To test a change, reproduce an incident or prep a demo, without building a local environment.
- No imposed tasks
- Ready in seconds
For hiring teams
Enterprise Assessments
Ordered scenarios under a global timer, live supervision, a score hidden from the candidate, and an exportable report. Grading reads the real cluster state, not a quiz answer.
- Supervised sessions, exportable report
- Private scenarios, SSO, EU data residency
Five environments, provisioned on demand
The scenario picks the machine it needs, and you configure nothing. Four of the five give each machine its own kernel, behind hardware isolation.
Scroll sideways
Drive Klastro from your AI agent
Claude Code, Codex, claude.ai and ChatGPT connect to the Klastro MCP server and work under your account: write a scenario, start a lab, read a candidate’s report. Every tool is generated from the platform’s own API, so the agent can do what the app does, under the same rules.
Claude Code, in one command
claude mcp
add --transport http klastro https://mcp.klastro.io/mcp Codex, claude.ai and ChatGPT: the same URL. Sign-in goes through OAuth, with no key to paste.
- Claude Code
- claude.ai
- Codex
- ChatGPT
claude · ~/cks-prep
Turn our CKS prep notes into a graded challenge on kubeadm. Ten points a task.
-
klastro · list_environment_providers
kubeadm-single · own kernel, hardware-isolated
-
klastro · create_studio_draft
Draft “CKS Custom Scenario”
-
klastro · update_studio_draft_graph
1 Helm release · 6 shell scripts · 14 tasks · 140 pts
-
klastro · validate_studio_draft
No errors
-
klastro · start_studio_preview
Preview ready on kubeadm-single
-
klastro · test_task_validation
Every check fails on the untouched cluster, as it should
The draft is in Studio, unpublished. Review the canvas, then publish.
- Author
- Studio drafts, a preview on a real cluster, checks tested before anything is published.
-
- create_studio_draft
- start_studio_preview
- publish_studio_draft
- Practise
- Start a lab, ask for a hint, submit a task, pause and resume.
-
- create_session
- reveal_hint
- submit_task_attempt
- Assess
- Set up an assessment, follow its status, read and export a candidate’s report.
-
- create_assessment
- get_candidate_report
- export_report
The agent acts with your permissions and nothing more. Billing, administration and your profile are not exposed to it at all.
You need a Klastro account to connect. Request an invitation
Three self-serve plans, and a contract for enterprises
A real lab burns real resources. Lab time is metered in credits, and no overage is ever invoiced.
Community
Free
10 credits, granted once
Enough to touch real infrastructure and find out whether the format suits you.
- Tutorials, challenges and playgrounds
- Community scenarios from the catalogue
- Welcome credits go to the founder once and are never renewed
- Invited members start with no credits
Pro
€25 per seat, per month
or €250 per seat per year, which is two months free
60 credits per seat, per month
The full catalogue, for every member holding a seat.
- Everything in Community
- Every community and Pro scenario
- Seats your admins assign and reassign across the team
- Credits reset to the grant each month; unused credits do not roll over
Instructor
€50 per seat, per month
or €500 per seat per year, which is two months free
100 credits per seat, per month
Everything in Pro, plus what it takes to author scenarios and run sessions.
- Everything in Pro
- Scenario Studio: author, preview and publish your own scenarios
- Instructor-led cohorts, built from reusable tracks
- 3 cohort places included per cohort
- Give credits from your personal balance
On every plan
- Labs draw 1 to 6 credits an hour, depending on the environment
- Starting a lab needs one hour of its credits in your balance
- Labs pause when credits run out. Paused time is not charged, and no overage is invoiced
- One running lab per scenario for each member
- Any member can redeem a credit voucher
Enterprise Assessments
CustomA contract, set up with our team rather than bought online. It is the only offering that opens proctored hiring assessments.
- Proctored assessments, with scoring and reports only your recruiters see
- Graded by validation rules against real cluster state, reproducible task by task
- Every scenario for every member, with no seats to assign
- Scenario Studio, SSO and private scenarios
- Concurrent lab capacity agreed in your contract
- Per-session network isolation, data hosted in the EU
These are the prices the app charges. During the private preview, sign-up is by invitation.
See the platform on a real hiring case
A walkthrough with the team: an assessment scenario opened in front of you, live supervision, and the report a recruiter receives. Pick a slot.
The calendar is provided by Cal.com. It loads only once you click, and nothing reaches that third party before then.
Calendar not showing? Open the booking page on cal.com
Common questions
What people ask us most.
- Is Klastro open?
- Yes, as a private preview. The platform runs in production and accounts open by invitation, in waves, so every newcomer gets labs that actually start. Ask for an invitation and tell us what you want to do with it.
- How do I get an invitation?
- Email us and we answer every message. No form, no tracker, no mailing list: tell us what you are trying to do. The invitation arrives by email with the next wave.
- What exactly is a credit?
- A credit is lab time. The hourly cost depends on the environment: a vcluster burns less than a two-node kubeadm cluster, because it uses fewer resources. On self-serve plans, your balance is visible at all times in the dashboard.
- Do I need to install anything?
- No. The terminal, the editor and the instructions all open in the browser. You need a connection and nothing else.
- Can I use Klastro from Claude Code or Codex?
- Yes, through the Klastro MCP server. Your agent signs in with your account over OAuth and gets the same permissions you have in the app: it can write a Studio scenario, start a lab or read a report, but has no access to billing or administration.
- Where is my data hosted?
- In the European Union. Session clusters, the database and the backups all stay in the EU, and every session gets its own isolated network space.
- Is this suitable for CKA, CKAD or CKS prep?
- It is one of the uses Klastro is built for. Scenarios run on real clusters and are graded against actual cluster state, the way a practical exam works, and timed challenges reproduce the time pressure. Klastro prepares you for those exams; it is not affiliated with or endorsed by the Linux Foundation.
Ask for an invitation
Klastro is in private preview. Tell us in one line what you want to do with it. We answer everyone, and the invitation follows with the next wave of accounts.
or write to us directly at [email protected]
An email that lands in a real inbox. We answer every one.