Introduction - Ankra Platform

Are you an AI agent or automating Ankra? Don’t drive the web UI with a browser. Connect over the MCP server (recommended) or the CLI - both give you the same tools the product’s own AI uses, with scoped read/write access and a full audit trail. Your CI jobs can call the CLI directly too - see Pipeline Agents. Setup is in Connect an AI agent below.

Ankra Platform Overview

What is Ankra?

Ankra is one platform to build, deploy, and debug Kubernetes - operated from whichever surface fits the job: the web UI for humans, the CLI for terminals and scripts, an AI agent over MCP for Claude, Cursor, and other MCP clients, or your CI/CD pipelines. Import any cluster - EKS, GKE, AKS, on-prem, or local - and manage it from one place. Or provision managed clusters on Hetzner, OVHCloud, or UpCloud directly from the platform. All four surfaces share one tool registry: the same 169 tools power the in-product AI Assistant, the MCP server, chat over Slack and Microsoft Teams, and SCM comments - so what you can ask Ankra to do is the same everywhere.

Four Ways to Operate Ankra

Web UI

The visual Stack Builder, the ⌘+J AI Assistant with Ask, Agentic, and Plan modes, and a full Kubernetes resource browser. Built for humans.

CLI

ankra chat "..." --mode ask|agent for one-shot AI turns, ankra cluster apply -f for declarative cluster operations. Built for terminals, scripts, and CI.

MCP for AI Agents

A remote MCP server exposing 169 tools with mcp:read/mcp:write scopes and OAuth. Built for Claude, Cursor, and any other MCP client.

CI/CD & IaC

GitHub Actions and GitLab image-bump pipelines, Pipeline Agents (plain CI jobs that call the CLI), and a Terraform provider. Built for your delivery workflows.

Connect an AI agent

Ankra is built to be operated by agents, not just people. No browser needed: don’t scrape the platform or automate the web UI - connect over MCP or the CLI instead. Fastest path for Claude Code - register the MCP server once, then authenticate with /mcp:

claude mcp add --transport http ankra https://platform.ankra.app/api/v1/mcp

For Claude desktop/web add Ankra as a custom connector; for Cursor drop the endpoint into your MCP config. Full setup for every client, token scopes, and the OAuth flow are in the MCP Server guide. Then teach your editor’s agent Ankra’s conventions - ankra skills install installs Ankra’s curated Agent Skills into Cursor (default) or Claude Code (--editor claude-code).

MCP Tool Reference

Every tool the MCP server exposes, the scope it needs, and whether it targets a specific cluster.

AI Skills

Teach Ankra’s AI your conventions with organisation-scoped skills that override the platform defaults.

Pipeline Agents

Plain CI jobs that call ankra chat and ankra cluster apply - PR review, deploy watching, health checks. No framework to run.

Hermes

Ankra’s autonomous infrastructure agent, operated from Slack, Discord, or Telegram chat.

Start read-only (mcp:read) to explore, then grant mcp:write when the agent should apply changes. Every resolved tool call - including refused ones - is written to the organisation audit log.

Build Your Stack in the UI

Stop copy-pasting YAML between clusters. The Stack Builder gives you a visual canvas to compose Kubernetes environments that are reusable, versioned, and deployed consistently.

Visual drag-and-drop - add Helm charts and manifests, define dependencies, configure values
AI-assisted composition - describe what you need (“monitoring stack with Prometheus, Grafana, and alerting”) and the AI recommends components, configurations, and deployment order
Variables - parameterise manifests with organisation, cluster, and stack-level variables using ${{ ankra.variable_name }} syntax
Clone across clusters - replicate a stack from dev to staging to production in seconds
SOPS encryption - encrypt secrets in your GitOps repository with built-in SOPS support

Automate Through GitOps and CI/CD

Ankra’s GitOps engine is designed to slot into your existing workflows, not replace them. Connect a GitHub repository, and your cluster configuration becomes code - versioned, reviewable, and automatically applied.

Works with existing pipelines

Your CI builds and pushes container images. A commit to the GitOps repo updates the image tag. Ankra detects the change and deploys. No new tooling required.

Modular repository structure

Split your configuration into files and folders with include paths. Different teams own different files. No merge conflicts, no giant config files.

apiVersion: v1
kind: ImportCluster
metadata:
  name: my-cluster
spec:
  git_repository:
    provider: github
    credential_name: my-credential
    branch: main
    repository: my-org/my-repo
  stacks:
    - name: platform-stack
      manifests:
        - include: manifests/
      addons:
        - include: addons/

Full audit trail

Every sync is tracked - commit SHA, trigger source, status, and timing. View sync history, monitor progress, and debug failures from the platform or via the API.

Plug in your pipelines

GitHub Actions Pipeline

The canonical build → push → image-bump → sync pipeline for GitHub repositories.

GitLab Pipeline

The same pipeline for GitLab CI, using the GitLab Container Registry and CI variables.

Pipeline Agents

Turn plain CI jobs into infrastructure agents with ankra chat and ankra cluster apply --wait.

Applications (Closed Beta)

Repo in, generated Dockerfile, Helm chart, and CI/CD out - shipped as a stack.

Prefer infrastructure-as-code? The Terraform provider manages clusters, stacks, manifests, and add-ons declaratively.

Manage and Debug with AI

Ankra’s AI isn’t a generic chatbot bolted onto a dashboard. It has full context - your pod logs, Kubernetes manifests, stack deployments, resource states, events, and their relationships. This makes it exceptionally good at incident triangulation.

AI Assistant

Press ⌘+J from anywhere. Choose Ask (read-only), Agentic (confirm each change), or Plan (approve once, then execute). The AI is page-aware - open a crashing pod and it already sees the logs, events, and manifest. Ask it anything:

“Why is this pod failing?” - correlates logs, events, and resource state
“Create a monitoring stack with Prometheus and Grafana” - designs and deploys a complete stack
“What changed in the last hour that could cause this?” - cross-references deployment history with symptoms

Proactive AI Insights

AI Insights scan your clusters on a schedule and surface issues before they become incidents:

Root cause analysis with remediation commands
Severity tracking and trend analytics (MTTR, category breakdowns)
Adaptive scanning - critical issues scanned every 60s, healthy clusters every 10 minutes

AI Incidents

When alerts trigger, the AI automatically collects pod status, events, logs, and node state - then delivers a structured analysis with root cause, affected resources, and recommended actions.

Everything Else

Resource Browser

Browse pods, deployments, services, secrets, RBAC, storage, CRDs, and Helm releases. Tail logs. Edit configs. No kubectl required.

Alerts & Webhooks

Configure alerts with AI-powered incident analysis. Send notifications to Slack, webhooks, or any external system.

Managed Clusters

Provision K3s clusters on Hetzner, OVHCloud, or UpCloud with automated networking, bastion access, and NAT gateways.

Cloud Cost

Estimated spend per cluster and across your fleet, broken down by compute, storage, and namespace.

Stack Profiles

Capture a working stack as a reusable, versioned, parameterised template and roll it out anywhere.

kubectl Access

Give teammates scoped, SSO-backed kubectl access through Ankra’s proxy - even to private clusters.

Fleet Dashboard

A world map of every cluster with health and cost rollups across your whole organisation.

API Reference

Full REST API for programmatic access to clusters, stacks, insights, and operations.

Quick Start

Create an account at platform.ankra.app

Add a cluster

Import an existing cluster by installing the Ankra agent, or provision a new one on Hetzner, OVHCloud, or UpCloud.

helm upgrade --install ankra-agent oci://ghcr.io/ankraio/ankra-agent/ankra-agent \
  --namespace ankra \
  --create-namespace \
  --set config.token="YOUR_UNIQUE_TOKEN"

Build your first stack

Open the Stack Builder or press ⌘+J and tell the AI what you need: “Set up a monitoring stack with Prometheus, Grafana, and Loki”

Connect GitOps

Link a GitHub repository in cluster settings. Your stacks are now version-controlled and automatically synced.

Full Getting Started Guide →

Next Steps

Getting Started

Step-by-step guide from first cluster to full GitOps pipeline.

Pipeline Agents

Turn plain CI jobs into infrastructure agents with the Ankra CLI.

Join the Community

Get help and share feedback.

​What is Ankra?

​Four Ways to Operate Ankra

Web UI

CLI

MCP for AI Agents

CI/CD & IaC

​Connect an AI agent

MCP Tool Reference

AI Skills

Pipeline Agents

Hermes

​Build Your Stack in the UI

​Automate Through GitOps and CI/CD

​Works with existing pipelines

​Modular repository structure

​Full audit trail

​Plug in your pipelines

GitHub Actions Pipeline

GitLab Pipeline

Pipeline Agents

Applications (Closed Beta)

​Manage and Debug with AI

​AI Assistant

​Proactive AI Insights

​AI Incidents

​Everything Else

Resource Browser

Alerts & Webhooks

Managed Clusters

Cloud Cost

Stack Profiles

kubectl Access

Fleet Dashboard

API Reference

​Quick Start

​Next Steps

Getting Started

Pipeline Agents

Join the Community

What is Ankra?

Four Ways to Operate Ankra

Connect an AI agent

Build Your Stack in the UI

Automate Through GitOps and CI/CD

Works with existing pipelines

Modular repository structure

Full audit trail

Plug in your pipelines

Manage and Debug with AI

AI Assistant

Proactive AI Insights

AI Incidents

Everything Else

Quick Start

Next Steps