Klastro and AI assistants (MCP)
Klastro has a Model Context Protocol (MCP) server. MCP is an open standard that lets AI assistants such as Claude, Cursor or GitHub Copilot in VS Code call outside tools. Once you connect an assistant to Klastro, you can ask it in plain language to do things you would otherwise do in app.klastro.io.
The server lives at:
https://mcp.klastro.io/mcpWhat you can do with it
Section titled “What you can do with it”The assistant works with the same features you see in the app, for the organization you choose when you connect it:
- Browse the catalog: list scenarios and read their details.
- Practise: start a lab, pause, resume or end it, submit a task for grading, reveal a hint, keep notes, and review your challenge history.
- Follow learning paths: see your progress and mark steps as started or complete.
- Run hiring assessments (recruiters): create and edit assessment templates, follow candidates’ progress, authorize a start, adjust the time, end an assessment, read candidate reports, export them as PDF and submit your review.
- Author scenarios (Scenario Studio): create and edit drafts, validate and publish them, start a live preview, push changes to it, and test a task’s checks before learners see them.
For example, you can ask: “List the Kubernetes scenarios available to me”, “Which candidates finished the Platform Engineer assessment this week, and what did they score?”, or “Start a preview of my draft and test the checks of task 2”.
The full list is in the tools reference. Billing, organization administration, invitations, legal documents and profile settings are not available through MCP; use the app for those.
How access works
Section titled “How access works”You sign in with your Klastro account
Section titled “You sign in with your Klastro account”There is no API key to copy. The first time your assistant calls Klastro, it opens your browser on the Klastro sign-in page. You sign in as usual, then a consent screen asks you to approve the connection. The assistant never sees your password.
You choose the organization and the permissions
Section titled “You choose the organization and the permissions”The consent screen, titled after the application (for example “Claude Code wants to access your Klastro account”), asks you to choose:
-
Organization: the application will act only in this organization. To work in another one, connect again and pick it.
-
Permissions:
Permission What it allows Read View scenarios, sessions, assessments and reports. Write Start and stop labs, edit Studio drafts, submit attempts. Administration Manage the organization. Admins only. Never ticked by default.
Select Allow to approve or Deny to refuse. A connection with only Read can look at things but cannot change anything: every action that starts, edits, submits or deletes needs Write.
Your own rights still apply
Section titled “Your own rights still apply”The assistant acts on your behalf, with your rights. Granting a permission never gives it more than you have in the app: a learner’s assistant cannot read candidate reports, and an assistant cannot open Scenario Studio for someone who has no Studio access.
Credits are spent like in the app
Section titled “Credits are spent like in the app”Starting or resuming a lab through an assistant costs the same credits as doing it in the app, and the same limits apply (balance, sessions running at once). Studio previews do not consume credits. See Plans and credits.
You can revoke access at any time
Section titled “You can revoke access at any time”Your Profile page lists every application you have authorized under Connected applications, with its permissions and organization. Select Revoke next to one to cut its access immediately: its tokens stop working and it has to ask for your authorization again.
Usage limits
Section titled “Usage limits”Each user can make up to 120 requests a minute through the MCP server. Beyond that, requests are refused for a short time and the assistant should retry later.