Finance

How Nat Eliason's OpenClaw Bot Felix Runs Its Own Business

Built by Nat Eliason · · Source: YouTube

Quick answer

Nat Eliason gave his OpenClaw bot, Felix, $1,000 plus access to Vercel, Stripe, GitHub, an X account and a crypto wallet, and told it to build a business. Felix built and launched a website and a paid PDF guide overnight, and handles most of its X replies on its own.

Result: $14,718 made in 3 weeks from a $1,000 start, according to the video description

What they built

Nat Eliason runs an OpenClaw bot called Felix on its own Mac Mini. In an interview with Peter Yang, Nat says Felix's task is to start a million dollar business that it runs itself: building the products, doing support, doing marketing and posting about it on X.

The first test was an overnight one. Nat says they told Felix it had Vercel access and Stripe keys, asked for a product it could build entirely on its own, and went to sleep. By morning, according to Nat, Felix had made a website (felixcraft.ai), written a PDF guide on getting a basic OpenClaw bot up to Felix's level, and connected Stripe to the landing page. The only thing it needed from Nat was the DNS settings. Felix then launched the guide on X.

How it works

  • Telegram threads. Nat talks to Felix in a Telegram group chat with a separate topic for each project, such as the X account, an iOS app, a document editor called Polylog and a hosted version of Felix called EasyClaw. Nat says each topic starts its own OpenClaw session, so Felix can work on several things at once without the contexts mixing.
  • X account. Nat says Felix handles 90 to 95% of its X activity automatically, especially replies. New tweets go to Nat for approval first, and Nat says they mostly approve them unedited.
  • Security. According to Nat, OpenClaw separates authenticated command channels from information channels. X mentions and email count as information, so Felix ignores prompt injection attempts there. Only Nat's phone on Telegram and the Mac Mini itself can give it commands. Whenever Nat adds a permission, they ask Felix how to lock it down before using it. Nat also warns people not to treat this setup as proven safe.
  • Crypto wallet. Someone else launched a Felix token, and Nat receives a share of its trading fees. Nat set up a daily automation that claims the fees, burns half of the Felix tokens and sends the rest to Felix's own wallet.

The video also covers a three-layer memory system, heartbeats, cron jobs and delegating to Codex, but the transcript we have cuts off before those details.

The result

Four days after the launch, Nat showed a Stripe dashboard with $3,596 in gross volume and $3,440 net from the PDF. The video description says Felix has made $14,718 in the three weeks since Nat gave it $1,000. Nat also says Felix's X account had about 2,500 followers and its crypto wallet held roughly $80,000 at the time of recording, down from about $100,000 earlier.

How to build something similar

  • Start with one narrow task that has a clear finish line, like a landing page with a payment link, before handing an agent a whole business.
  • Give the agent its own accounts and scoped API keys, and add access one piece at a time so you can see what each new permission changes.
  • Split work into separate chats or sessions per project so one task doesn't clutter the context of another.
  • Keep a human approval step on anything public or involving money until you trust the output, and treat everything the agent reads online as information, never as instructions.

The original post

Peter Yang · February 22, 2026 · View on YouTube

Want to build something like this?

Join the Shipping Skool community free. Paid members get live bootcamp calls every week.

Join Shipping Skool — Free

More use cases

← All AI agent use cases