Show HN: A platform to run Docker container jobs and host always-on servers (mermadicrunner.com)

I built Mermadic because I was tired of the usual split: either a shared Heroku/Render-style box (fine until noisy neighbors, sleep, or a surprise bill), or standing up EKS and suddenly Kubernetes is a second job.

You can use default cluster to run jobs or subscribe for isolated infrastructure for container jobs and always-on APIs. Dedicated is your own VPC, cluster, and mermadic control plane API, not a namespace on a shared cluster. This is a really simple platform, you don't operate Kubernetes and don't need an AWS account.

What you can do: - One-off Docker jobs and cron (bring a Docker Hub image, or build one in our CDCI pipeline) - Always-on deployments with HTTPS and custom domains (dedicated only) - GitHub: push a branch, auto build and can redeploy if configured - REST API, or a Cursor/Claude skill that runs jobs/deploys from chat - MCP server (mcp.mermadicrunner.com) so an agent can sign up, add credit, push code and run a job

default is free and multi-tenant. Jobs and cron only — good for kicking the API. When you are in default cluster, you can click through any features to see how the workflow is like. Dedicated clusters is when you want isolation and something that stays up. Jobs cap at 24 hours. Hosted, not open source. If you already happily run your own cluster, then this is probably not needed.

Fastest try: sign up, My Account, Developer Access, create a key, then:

curl -X POST https://job.mermadicrunner.com/job-runner/v1/dockerhub/creat... \ -H "Authorization: Bearer YOUR_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "image": "python:3.11", "commandArgs": ["python", "-c", "print(\"hello from mermadic\")"], "resourceRequests": { "cpu": 0.5, "memory": 0.5 } }'

Docs: https://www.mermadicrunner.com/documentation

I'm here for the next few hours, please let me know what you think of this or is there any advice.

If you were moving a real app off Heroku or Render, what would make isolated infra actually useful vs just another PaaS? Anything in here that feels like the wrong abstraction? And if you live in Cursor, is MCP the interface you'd want, or would you rather the agent just hit a normal job API?