The Twitter scheduler category is a graveyard. He didn't win by building a better one. He won by shipping the plumbing a specific tribe needed the exact week they started searching for it.
The case at a glance
- Product: OpenTweet, a Twitter/X scheduler that doubles as an MCP server for Claude, Cursor, and OpenClaw agents
- Price: $11.99/mo, no annual push, ~$13.50 blended ARPU
- Dec 2025 MRR: $12 (one paying customer)
- Feb 2026 MRR: $230, near quit point
- Mar 2026 MRR: $2,200 — MCP feature ships, roughly 10x in one month
- May 2026 peak: $4,200 MRR; stable at $3,400 MRR today with 251 paying subs
- Distribution: ~80% SEO, ~20% Reddit and Twitter organic, $0 ad spend
Stack: Next.js 14, TypeScript, MongoDB, Auth0, Docker Compose on a single $12 DigitalOcean droplet. One person, Balkans-based, out-earning his DevOps day job by roughly 3x after infra and tax.
What he's actually selling: the MCP server, not the scheduler
The product isn't a Twitter scheduler. The product is price arbitrage against a $5,000/month API tier.
X charges $5K/mo for the API level his customers need. He charges $11.99. That's not a discount, it's a 400x gap. Every AI-agent builder who wants to post to X programmatically has two real options: his tool, or nothing. The category has fifty Twitter schedulers and none of them went $12 to $3.4K in seven months, because none of them are actually competing in the same market he is.
The scheduler UI is the storefront. The MCP server, REST endpoint, and CLI are the actual product. Sold naked to developers, the plumbing would be less discoverable, less billable, less defensible. Wrapped behind a consumer surface, it ranks for "Twitter scheduler" on Google while also getting picked up inside Claude Desktop and Cursor for "twitter MCP server". Two products, one codebase, one droplet.
Don't sell what the category calls itself. Sell what the customer is actually renting.