<!-- Generated from the canonical HTML page. Do not edit separately. -->
# An AI platform engineer - Ankra Use Cases

Canonical: https://ankra.ai/use-cases/ai-platform-engineer

Agents that read live cluster state, fix what breaks and wait for your approval. Use Ankra's AI team, or connect Claude, Cursor or your own agent over MCP.

[All use cases](https://ankra.ai/use-cases)

An AI platform engineer

# Put an AI platform engineer on every cluster.

Agents that read live cluster state, fix what breaks and wait for your approval. Use Ankra's AI team, or connect Claude, Cursor or your own agent over MCP.

[Start with Ankra](https://platform.ankra.app/auth/signup)  [Read the guide](https://docs.ankra.ai/platform/ai-agents)

Connect Claude Code to every cluster in one line

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

Run `/mcp` in Claude Code to sign in. Cursor and other clients use a token.  [MCP docs](https://docs.ankra.ai/platform/mcp-server)  [CLI docs](https://docs.ankra.ai/integrations/ankra-cli)  [CLI on GitHub](https://github.com/ankraio/ankra-cli)

For engineering teams and agent builders

The work to take off your team

## An agent can write a fix. It needs the running system to know which fix matters.

Ankra brings cluster state, logs, events, Git configuration and operational tools into the same platform. Use the built-in AI team or connect your existing agent through the CLI, API and MCP.

With Ankra

### Agents that investigate, propose the fix and act only within the boundaries you choose.

Choose the agent  Grant useful context  Set the boundaries  Follow the work

A practical path

## How this works with Ankra.

1. 01

  ### Choose the agent

  Use a role agent, create your own, or connect the agent you already work with.
2. 02

  ### Grant useful context

  Scope its access to the organisation, clusters and individual tools it needs.
3. 03

  ### Set the boundaries

  Choose observation, proposals or permitted autonomous actions. Review gated changes before they run.
4. 04

  ### Follow the work

  Inspect runs, transcripts, board tickets, approval requests and execution records.

 [Follow the product guide](https://docs.ankra.ai/platform/ai-agents)

Hand it real work

## Ask it like you would ask your best platform engineer.

Every answer starts from the running system: logs, events, manifests, Git history and cost. Every change runs inside the boundary you set.

- Read only

  “Why is checkout-api crashlooping in production?”

  Reads the previous container's logs, the events for that pod and the stack's recent changes, then answers with the evidence.
- Waits for approval

  “Upgrade cert-manager on staging, then hold production for me.”

  Proposes the chart upgrade with the current version as the rollback and parks the production change until you approve it.
- Pull request is the gate

  “Clear the critical CVEs on prod-eu.”

  Maps each vulnerable image to the application or add-on that deployed it and opens one pull request per repository with the fixed versions.
- Read only

  “What is this cluster costing us, and where is it idle?”

  Pulls estimated cost by cluster, stack and workload, and points at the unused capacity worth reclaiming.
- Pull request is the gate

  “Put this repository online on staging.”

  Analyses the repository and proposes the Dockerfile, chart and pipeline it needs as a pull request you merge.
- Read only

  “Is anything degraded across my clusters right now?”

  Checks every cluster in your organisation and summarises what needs attention first, with portal links to each one.

Bring your own agent

## One command.  Every cluster behind it.

Claude, Cursor or the agent you built yourself gets the same tools Ankra's AI team uses. There is nothing to install or host: point the client at the endpoint, sign in, and ask for your clusters.

- [Ankra CLI docs](https://docs.ankra.ai/integrations/ankra-cli)
- [ankraio/ankra-cli, open source on GitHub](https://github.com/ankraio/ankra-cli)
- [MCP server setup for Claude, Claude Code and Cursor](https://docs.ankra.ai/platform/mcp-server)
- [Every MCP tool and the scope it needs](https://docs.ankra.ai/platform/mcp-tools)

# Register the server, then sign in with OAuth

```
$ claude mcp add --transport http ankra \
    https://platform.ankra.app/api/v1/mcp
$ claude
> /mcp                # sign in to Ankra
> why is checkout-api crashlooping in production?

  ankra · list_clusters
  ankra · get_pods        namespace checkout
  ankra · get_pod_logs    previous container
  ankra · get_events      pod checkout-api
```

Illustrative session. Claude's OAuth grant is scoped to one organisation.

# Install, sign in and teach your assistants Ankra

```
$ brew install ankraio/tap/ankra
$ ankra login
$ ankra skills install   # Agent Skills for every assistant found
$ ankra chat "what is degraded across my clusters?"

# Machine-readable output for scripts and CI agents
$ ankra chat --mode ask -o json "is anything degraded?"
$ ankra agents runs
```

No Homebrew? Use the install script or a binary from GitHub Releases.

# A read-only token, then your user-level MCP config

```
$ ankra tokens create cursor --scopes mcp:read

# ~/.cursor/mcp.json
{
  "mcpServers": {
    "ankra": {
      "url": "https://platform.ankra.app/api/v1/mcp",
      "headers": { "Authorization": "Bearer <ankra-token>" }
    }
  }
}
```

Replace <ankra-token> with the token the first command prints. Add mcp:write only when the client should change resources.

MCP tools across clusters, stacks, GitOps, pipelines, security and cost

450+  MCP tools across clusters, stacks, GitOps, pipelines, security and cost

role agents in every organisation from day one

6  role agents in every organisation from day one

token scopes. mcp:read looks, mcp:write acts

2  token scopes. mcp:read looks, mcp:write acts

worker vCPUs free on every plan, try it without a card

30  worker vCPUs free on every plan, try it without a card

Give your agent a cluster tonight. Start free, connect a cluster and ask your first question in minutes.

[Start free](https://platform.ankra.app/auth/signup)  [Try a playground first](https://ankra.ai/playground)

Already on staff

## Your AI team is waiting when you sign up.

Six role agents arrive in every organisation. Chat with one, dispatch it on a mission, or build your own with the agent wizard.

- ### AI Team Lead

  Triages requests, routes work to the right specialist and files tickets for anything that outlives the chat.
- ### Developer

  Helm charts and values, deploys, CI/CD workflows and pull request demos.
- ### Database Admin

  StatefulSets, volumes, operators and backups. Verifies backups before proposing anything that touches data.
- ### SRE

  Alert triage, golden signals, incident timelines and postmortem notes. Evidence first.
- ### DevOps

  Clusters, node pools, add-ons, stacks, GitOps drift and IaC. One reversible step at a time.
- ### DevSecOps

  Turns a security report into fixes, as pull requests at the source and add-on upgrade proposals.

Safe by default. They never run on a schedule, every cluster write waits for your approval, and they cost nothing until you talk to them.

The platform behind the workflow

## The pieces, already connected.

[### Ankra AI agents Work with role agents for delivery, reliability, databases, infrastructure and security, or create an agent for your team.](https://ankra.ai/features#agents)

[### CLI, API & MCP Connect the agent you already build with to the same platform your team uses. Automate from your editor, terminal or scripts.](https://ankra.ai/features#mcp)

[### Agent autonomy & approvals Choose observation, proposals or permitted autonomous actions. Review gated changes and follow what agents executed.](https://ankra.ai/features#approvals)

Watch a real workflow

## See it running.

A recorded walkthrough of the platform on real infrastructure. The film shows a specific environment, rather than a promise about every deployment.

Plays from YouTube.  [Open on YouTube instead](https://www.youtube.com/watch?v=XCgtWyLb-5o)

What comes next?

## Keep building from here.

[Operate & protect ### Investigate production incidents Give your team and its agents the logs, events and resource state needed to investigate a running application.](https://ankra.ai/use-cases/incident-response)

[Operate & protect ### Self-service with control Give developers and agents useful access while retaining ownership of permissions, configuration and changes.](https://ankra.ai/use-cases/team-governance)

[Build & ship ### Application delivery & CI/CD Bring testing, builds, security scans, images and deployment into one delivery workflow.](https://ankra.ai/use-cases/ci-cd-pipelines)

Your next chapter

## You built it.
*Let’s get it live.*

Give your application a path to production.
 Give your team a simpler way to run it.

[Get your app live](https://platform.ankra.app/auth/signup)  [Try a playground](https://ankra.ai/playground)

First 30 vCPU of platform usage included.
Try it without a card; a card unlocks full access and €59 of credit.

From something you built
to something people use.

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