Alpha: AWS, self-service

Production readyin minutes.

Deploy now
Connect your agents

A cursor drives the real UI — eight services, two databases and their wiring into a customer AWS account. Loops about every 25 seconds.Open it full-bleed

Shipping on their own cloud accounts

  • Northwind
  • Lumen Health
  • Vantage
  • Cobalt
  • Meridian
  • Trellis

How it works

Three steps to your first deploy

  1. Connect your account

    Run one CloudFormation stack — or `terraform apply`, if that is your religion — to create the Kapten cross-account role. It grants scoped provisioning rights and nothing else. About two minutes.

    aws
    $ aws cloudformation deploy \    --stack-name kapten-role \    --capabilities CAPABILITY_NAMED_IAM  CREATE_COMPLETE  kapten-role

  2. Provision an environment

    Pick a region and hit go. Kapten lays down the private network, the compute, the routing and the dashboards — all on current best practice. You watch it happen; you approve before anything is applied.

    kapten
    Cloud
    AWS · eu-west-3
    Environment
    production · 3 AZ
    Includes
    routing · TLS · observability

    Awaiting approval — nothing applied

  3. Deploy your blueprint

    Describe your app once — its container, its database, its cache, its hostname. Kapten builds the image in your account and rolls it out behind health checks. Every environment after the first is the same blueprint, one click away.

    git
    $ git push origin main  build     ✓ 41s → payments:9f2c1ab  rollout   ✓ 3/3 instances healthy  live      https://payments.acme.dev

Bring your own cloud

Your account. Your network. Your bill.

Managed platforms keep your data in their account, their region, on their invoice. Kapten inverts that.

Kapten holds

  • Deploy metadata
  • Log index
  • Nothing else
  • No data leavesmetadata only
  • Committed spenddiscounts apply
  • Leave wheneverplain Kubernetes
  • Scoped access900-second sessions

Scaling

Ten times the traffic. The same Tuesday afternoon.

Capacity follows load on infrastructure that is already running. No new environment, no migration, no window. You keep shipping, and nobody gets paged.

Set a floor and a ceiling per service. Kapten stays inside them.

Requests / sec1,240p95118 msflat
Load balancerHTTP/3×1apinode 20×2workerjobs×1postgres15.4×1cache7.2×1

Steady at api ×2. Nobody was paged.

Agent-native

Talk to your infrastructure.

Kapten exposes your whole platform as an MCP server. Connect it once to the agent you already work in, then just ask. “Deploy payments to production”. It reads your environments, opens the deploy, watches the rollout and tells you where it landed.

  • Your role, not a superuser

    The connection inherits the exact permissions of the member who opened it. An agent cannot reach an environment its human cannot reach.

  • Every call is on the record

    Each tool call lands in the same audit trail as a dashboard action, attributed to the person behind the agent. Nothing happens anonymously.

  • Any MCP client

    Claude Code, Codex, Cursor, Zed. One remote endpoint, nothing to install and no local daemon to keep alive.

Endpointhttps://mcp.kapten.io/v1
claude · ~/acme/payments
Welcome to Claude Codekapten mcp server · 14 tools loaded
3 tool calls · scoped to your role

Ready when you are

Your cloud. Our ergonomics.

Connect a cloud account and watch a production-grade environment come up in your own VPC. If it isn’t serving traffic in fifteen minutes, we want to hear why.

Connect your cloud accountRead the docs