Ploink · Nov 21, 2026

A $10 GitHub Actions cap pushed this founder to deploy from his Mac

After Vercel overages and GitHub Actions minutes on about 15 side projects, the founder of Ploink built a one-button deploy console for the Mac.

The founder of Ploink set his GitHub Actions spending limit at a $5/month average. He hit it, raised it to $10, and thinks he is hitting that now too. That limit was the last step in a chain of workarounds across the roughly 15 side projects he runs, most of them built with AI agents.

Ploink is his answer: a Mac app that gives each project its own launch button. Tap F5 in any app, hold the button, and the deploy runs. Coding agents can ask to press it too.

From Vercel to a home lab runner to his own machine

The problem started with Vercel. The founder was on the $20 paid plan and running $10 to $20 in overages, "recently even more," though he notes not all of that came from deployments.

To cut the bill, he moved deploys to GitHub Actions, which triggered a run on a local runner in his home lab server. The runner did the build and uploaded to Vercel or Cloudflare, wherever each app was hosted.

That moved the cost rather than removing it. "Then I hit the limit in GitHub action minutes as well, so they are charging me every month," he said. His explanation for how it added up: AI agents build a lot and push a lot of code.

He started deploying manually from his machine instead, and lost track of what was deployed and how. Some projects the agent deployed on its own. For others, the agent asked him to run the command, which he attributes to the agents' security settings.

"I wished there existed an app where I could just summon it with one key and click to deploy a project," he said. He built it for himself, with the goal of making it look good and feel pleasant to use.

The agent can ask, or not

The homepage describes an agent asking to deploy, the button glowing, and the user holding it. Ploink includes an MCP server that lets agents trigger deploys. Whether an agent's request runs straight away or waits for a human hold is a setting in the app, and the user chooses.

The founder currently runs every project in the manual mode. His reason is the interface: the button press, the animation and the small UI interactions that play on each deploy. "It really makes me happy when I have to deploy something and press that button," he said.

That interface is also where most of the build time went. "Getting the UI feel just right took most time," he said. The other time sink was distribution: he tried to post Ploink on a subreddit for Mac apps, and the moderators kept refusing it under the subreddit's strict rules.

No cloud, on purpose

The feature he deliberately left out is any kind of cloud deployment. "This is anti-cloud lol," he said.

He acknowledges the appeal of pushing code and having the app deploy itself. "But I don't want to be paying $20-30 a month for that." For small micro-SaaS side projects, he considers cloud pipelines overkill. The homepage puts it as: no runners, no cloud, no YAML. The app also finds the deploy command for each project and keeps a receipt of everything shipped.

He is direct about the tradeoff. "Downside is that it won't really fit workflow of people who run agents in the cloud, but for me at this point in my life, this is perfect."

The pre-order terms

Ploink is sold as a founding pre-order: $19 once, against a $29 price after launch, with no subscription. The goal on the pricing page is 50 founders by 19 December 2026. If fewer than 50 sign up, everyone is refunded automatically and the core stays free. If there is no signed build by 31 January 2027, everyone is also refunded automatically. Founders get the unsigned beta now, Ploink Pro for all of 1.x, and a vote on which Pro features ship first. The code that touches the shell is public.

Ploink is for Mac developers who deploy several small projects from their own machine and want agents to ask before shipping. It is at https://ploink.dev/.

This story was written from a written interview with the founder, conducted by an AI interviewer and reviewed by the founder before publication.

See Ploink →