What’s new in Ankra

The launches, newest first. Every change, day by day, is in the full changelog.

  1. Onboarding · Playground · AI

    Four ways to start: your AI agent, a playground, your app or your own cluster

    After you sign up, Ankra opens on one question: where do you want to start? Pick the one that matches what you have today. The others wait on Home.

    • Connect my AI agent - add Ankra to Claude Code, Cursor, Codex or another MCP client, and let it do a first job. About a minute, no card.
    • Try a playground - a small Kubernetes cluster that Ankra runs for you, with a starter website online on it. Free for 24 hours, no card.
    • Put my app online - your app from a GitHub repository, live at an HTTPS address and built again on every push. It runs on a free playground.
    • Connect my own cluster - a Kubernetes cluster you already run, connected through the Ankra agent. It asks for a payment method, saving the card charges nothing, and the first 30 worker vCPUs are free.

    Every choice leads to the same portal. Compare them in Choose where to start.

  2. AI · Cost

    Your own AI keys: run chats and pull request reviews on your own account

    Add your own Anthropic or OpenRouter API keys to your profile, and the AI work you start runs on your account, not on your organisation’s allowance.

    • Chats and reviews - your AI chats, and the reviews of pull requests you open on GitHub and Bitbucket Cloud, use your keys.
    • Several keys, in your order - add more than one and they are tried in the order you set. A key that runs out is skipped.
    • The credits you already have - this includes the monthly API credits some Claude plans come with.
    • No surprises mid-conversation - nothing moves to another account while a conversation is running.

    Set it up in Your Own AI Keys.

  3. Organisation · Access · Cost

    Projects: give each team its own room inside your organisation

    An organisation can now be split into projects. A project holds its own clusters, applications, pipelines and cost. People, roles and the invoice stay with the organisation.

    • Nothing moves until you say so - everything you have starts in your Default project, and you move resources into a new project when you create one.
    • Members who see one project - invite a person as Project only and they see nothing else.
    • Cost by project - read what each team spends on its own.
    • Its own channel - a project’s notifications can go to its own Slack or Teams channel.
    • From the terminal - ankra project manages projects and --project scopes a command to one.

    Read Projects to set up the first one.

  4. AI · Pipelines

    Ankra AI fixes a failing pull request on its own branch

    When a pull request from a branch of your repository fails its build, its security gate or a test step, Ankra AI now fixes it where it is.

    • On the same pull request - the fix is a commit on that branch, and Ankra watches the run the commit starts. No second pull request to review.
    • Merge conflicts too - a pull request that conflicts with its base gets a merge commit that resolves it.
    • Forks get a diagnosis - a pull request from a fork is explained, not changed.

    More in When a pull request fails its checks.

  5. Pipelines · GitHub

    The merge train: every pull request is tested as it will merge

    Two pull requests can each pass on their own and still break the base branch together. Turn on a repository’s merge train and that stops happening.

    • Tested as it will land - Ankra tests each queued pull request squashed onto the base branch and the pull requests ahead of it, then merges it through GitHub when that run passes.
    • A failure leaves the train - the pull request that failed drops out and the ones behind it are tested again without it.
    • No rebuild after merge - every merge is a tree a run already passed, so a pipeline can publish the image that was tested instead of building it again.
    • Stacks ride together - a pull request based on another pull request’s branch can join while the one below it is in the train.

    The merge train is for repositories on GitHub. Read how it works.

  6. Backups · Stacks · CLI

    Backups: protect a stack, take restore points, restore it in place

    A stack with data in it can now be protected from the portal, the CLI and the API, in every organisation.

    • A vault you own - register a bucket you already have or let Ankra provision one. Backups go there.
    • Protection on a schedule - protecting a stack schedules its databases and named volumes into the vault.
    • Restore points you can read - take one whenever you like, and see what it carries and what it could not capture before you rely on it.
    • Restore in place - put a stack back from a restore point behind a typed confirmation, and follow every capture and restore as a run.

    The steps are in Backups and restore points.

  7. Cost · Multi-Cloud · FinOps

    Compare Providers: your cluster, priced on every cloud

    “Would this be cheaper somewhere else?” now has an answer built from your actual cluster rather than a pricing page. Compare Providers takes the node groups, control plane, volumes and load balancers you are really running and prices them on every provider Ankra can build on, side by side with today’s metered bill.

    • Like-for-like matching keeps dedicated cores dedicated, so the figure is one you could actually move to rather than a best case built from shared vCPUs.
    • Cheapest fit shows the other end of the range when shared cores are acceptable.
    • Line by line - the bill is broken down by what drives it, so you can see which machine class is the reason and what changes if you move.

    Find it under Cloud Cost on any cluster. The economics behind it are on the cost optimization page.

  8. Playground · Onboarding

    Playground: a real Kubernetes cluster in about two minutes

    You can now try Ankra on a real, writable Kubernetes environment without connecting a cloud account, installing anything, or talking to us first.

    • One click in the portal, or ankra cluster playground create from the terminal.
    • Ready in about two minutes - four provisioning steps, and our timed run clocked it at eighty-two seconds.
    • A starter app already live on its own hostname over HTTPS by the time you arrive.
    • Real in every way that matters - kubectl works, nodes and pods are genuine, and stacks you deploy go through the same pipeline your production deploys use.

    Nothing you build in it is playground dialect: the stacks are standard Helm charts and Kubernetes manifests, and they move to a cluster of your own unchanged.

    Start one from the Playground page.

  9. Pipelines · CI/CD · Security

    Ankra Pipelines: CI that builds, scans and publishes in your own cluster

    Connect a repository and Ankra runs its CI on a cluster in your organisation, from a pipeline file committed next to the code.

    • One file, .ankra/pipeline.yaml - the pipeline is in the repository, and a new application’s setup pull request writes it for you.
    • Builds without a Docker daemon - images build in a rootless BuildKit pod and are addressed by digest from then on.
    • Scanned and gated before anything is published - Trivy, Semgrep and Checkov scan the build, and a gate judges the findings against your organisation’s policy. A pipeline can tighten that policy and cannot loosen it.
    • What ships is what passed - the publish stage re-tags the one digest the gate approved.

    A pull request from a fork runs with restricted authority.

    More on CI/CD pipelines, and the setup is in the Ankra Pipelines guide.

  10. Security · CLI · Fleet

    Security Center: findings ranked by what is actually being exploited

    Every cluster ships with thousands of CVEs, and severity alone will not tell you which ones matter. The Security Center ranks them by evidence of real-world exploitation instead:

    • CISA Known Exploited Vulnerabilities first - findings on the KEV catalogue surface above everything else, with their remediation deadlines.
    • FIRST EPSS on every finding - the exploitation probability sits next to the CVSS score, so a critical nobody is exploiting stops outranking a high that everybody is.
    • The fixed version, per occurrence - not “upgrade something”, but which image and which package version closes it.
    • Scanner freshness per cluster - a cluster that stopped reporting three weeks ago cannot quietly look clean.

    One command covers the whole fleet, and the same data is in the CLI, the portal and your AI tools:

    ankra security overview
    ankra security findings --known-exploited
    ankra security finding <finding-id>
    ankra security clusters

    More on the Security Center, and on why findings are not decisions on the blog.

  11. CLI · CI/CD · AI

    Pipeline agents: CI-native infrastructure review, deploy watching, and scheduled health checks

    Build infrastructure agents from plain CI jobs and the Ankra CLI - no framework, no orchestration layer:

    • PR review agent - validates cluster-definition changes against the live platform and comments on the PR with what will actually change, including live capacity headroom checks.
    • Deploy watcher - ankra cluster apply --wait on merge, verifies rollout health, and posts a root cause to Slack when a deploy fails instead of a red X.
    • Scheduled infra watcher - runs every 30 minutes, stays silent while healthy, posts to Slack only when something degrades.

    ankra chat is one-shot and scriptable, with server-side access to cluster context: logs, events, manifests, and stack history. Read the guide.

  12. CLI · AI · Integrations

    Use the Ankra CLI as a Cursor infrastructure subagent

    Cursor is good at application code but loses context the moment a change crosses into Kubernetes, Helm, and CD pipelines. The Ankra CLI now works as an infrastructure subagent: Cursor delegates cluster-aware questions (“why is this pod crashing on staging?”) to ankra chat, which answers with grounded cluster state instead of guesses.

    Includes llms.txt support on the docs site so agents can self-serve CLI usage. Read the setup guide.

  13. AI · GitOps · Governance

    Draft-first AI: every AI change is a reviewable draft before it can touch production

    AI agents at other companies have wiped real production databases this year. Ankra’s answer is architectural: the AI can only produce drafts. Every generated stack, values change, or fix lands as a draft with a full preview (what changes, what restarts, what depends on it), requires explicit human approval, and commits through GitOps with an audit trail and one-click rollback.

    The AI gets the speed; Git keeps the authority. How it works.

  14. Stacks · Environments

This page lists the launches. Every fix and smaller change is in the full changelog, which also has an RSS feed. Deeper dives are on the blog.

The best way to know Ankra is to use it.

Connect a repository and a cluster and see your application run.

From something you built to something people use.

  • Deploy from your repository
  • Run on infrastructure you choose
  • Operate with your team and agents