The case at a glance
- Product: X/Twitter scheduler with analytics, best-time-to-post, AI idea generation, and an MCP endpoint
- Pricing: $11.99/mo, single tier
- Revenue: $3,400 MRR in June 2026, $4,200 peak in May
- Subscribers: 251 paying
- Time on product: 2–3 hrs/day, mostly support
- Traffic mix: ~80% Google + LLM referral, ~20% X/Reddit
Stack is one DigitalOcean droplet, Docker Compose, Postgres, Stripe, Cloudflare. The operator is a full-time DevOps engineer in the Balkans running the whole thing on the side of a day job. Revenue trajectory across seven months: $12 → $30 → $230 → $2,200 → $3,000 → $4,200 → $3,400. The step function in March is the whole story, and it's not the story you'd guess from the product page.
What he's actually selling: the MCP endpoint, not the scheduler
The product isn't an X scheduler. There are twenty of those. The product is the only X scheduler an AI agent can call.
In March he shipped a Model Context Protocol endpoint into an existing scheduler. That week, Claude users with MCP clients could draft, schedule, and thread posts from a chat window. They told each other. Two hundred of his 251 subs arrived in the six weeks after that ship, before any incumbent could respond.
Here's why the mechanism works. MCP isn't a feature, it's a distribution channel with zero saturation in 2026. Buffer can't ship it fast, they have a board. Typefully has a roadmap. He shipped it in a weekend from a droplet, and for six weeks he was the only agent-callable option in his category. Every Claude user who tried him told three others, and his programmatic SEO pages got indexed by Google and the retrieval layers behind Claude and ChatGPT. Same pages, two crawlers, one flywheel.
The moat isn't the product. The moat is being three months faster than the incumbent on the one integration nobody has shipped yet.