$3.4K MRR from 251 Users: The X Scheduler LLMs Now Recommend

The Case: $3.4K MRR, 2–3 hours a day, one $12/mo tier

What he does. Runs an X/Twitter scheduler with analytics, best-time-to-post, and AI idea generation at ~$12/mo to 251 subs.

Backstory. He's a full-time DevOps engineer living in the Balkans (avg salary €1,300–1,500). Shipped the app in 2024, sat under $250/mo for three months, wanted to quit. In March 2026 he plugged an MCP endpoint into the scheduler and Claude/agent users flooded in — the app went $230 → $2,200 in a single month.

Product. - Schedule X posts and threads - Impression + engagement analytics - Best-time-to-post recommendations - AI-generated post ideas - MCP endpoint so Claude and other agents can post through it

Pricing. $11.99/mo entry tier. 251 subs at ~$13.55 blended = $3,400 MRR, implying a light upgrade path.

Key numbers. - Dec 2025 revenue: $12 (1 customer) - Jan 2026: $30 - Feb 2026: $230 - Mar 2026: $2,200 (MCP launch) - Apr 2026: $3,000 - May 2026: $4,200 (peak) - Jun 2026: $3,400 (−19% off peak) - Paying subs: 251 - Time on product: 2–3 hrs/day, mostly support - Traffic mix: ~80% SEO + LLM referral, ~20% X/Reddit - Infra: 1 DO droplet, Docker Compose

Stack. Docker Compose, single DigitalOcean droplet, blue-green deploy by switching DNS, MCP endpoint, programmatic SEO pages, X + Reddit for social.

Why this works (and what most readers will miss)

1. The niche is already looking for the tool. X schedulers are a mature category. Buyers know they want it, know it costs $10–20/mo, and search for it with commercial intent. He didn't create a market. He ranked in one.

2. DevOps skill = 95% margin as a feature. One droplet, no Vercel, no Supabase, no Kubernetes. His infra cost at $3.4K MRR is maybe $40. Every managed-service SaaS at his scale runs 40–60% margin. He runs 95%. That's invisible to competitors and it's why he can survive $12 pricing.

3. He accidentally built the first agent-native tool in his niche. MCP wasn't a feature. It was a distribution channel. Every Claude user with an MCP client got a scheduler that worked from a chat window, and they told other Claude users. Nobody in the thread named this for what it actually was.

The Distilled Read

Ship the boring product in the boring category and win on execution. X schedulers are done. Buffer exists. Typefully exists. Twenty indie clones exist. He shipped anyway because a saturated niche is a validated niche — the buyer is educated, the price is known, the search volume is real. Pick a mature category and you skip 18 months of "does anyone want this." He spent that 18 months on the only thing that mattered: making the tool findable.

The template is one droplet, one docker-compose.yml, one pricing tier. Everything he does argues against optionality. No microservices. No free plan drowning support. No tiers of tiers. Two environments, DNS swap to deploy, done. If you can't describe your infra in one sentence at 251 customers, you're paying tax on complexity you don't need. The point isn't that Docker Compose is best. The point is that his stack fits in his head, so 100% of his brain is free for the product.

The unit economics only work because he is the ops team. Read this carefully: 251 users × $13.55 × ~95% margin ≈ $3,230 of net cash a month for 60–90 hours of work. That's $40–50/hr net, which in the Balkans is a wealthy person's rate and in San Francisco is a joke. This case is not portable to a Bay Area engineer on $200K comp — they'd need $15K MRR to feel it. Your geography is your pricing power. Live somewhere cheap and keep a Linux box alive and you're playing on easy mode. Otherwise you need 4x the MRR for the same lift.

Distribution came from two channels he treated as one. He posted on X and Reddit and called it "cringe." Fine. Real traffic was 80% Google + LLM referral. Here's what people miss: those aren't two channels, they're one channel with two indexers. His programmatic SEO pages fed Google and the retrieval layers behind ChatGPT and Claude. When a Claude user asked "best X scheduler with API" in June, the model surfaced a page he wrote in February. Every content operator in 2026 is writing for two crawlers now, whether they know it or not.

Here's the thing nobody in the thread saw. He thinks March clicked because "openclaw hyped people joined." That's the surface read. The real read: he shipped an MCP endpoint before any of his direct competitors, which meant for roughly six weeks he was the only X scheduler an AI agent could actually use. Buffer can't ship MCP fast — they have a board. Typefully has to prioritize. He shipped it in a weekend from a droplet. That six-week window is where 200 of his 251 subs came from, and he thinks it was luck. It wasn't luck. It was the compounding advantage of shipping speed in a category where the incumbents cannot respond in under a quarter. Eighteen months from now the winners in every indie category will be the ones who made their tool agent-callable first. The corpses will be the ones still building nicer UI for humans.

Steal-the-Playbook

1. Pick one saturated micro-vertical, not the category. Not "creator tools." Not "social media." Pick "X schedulers with analytics," or "LinkedIn carousel makers," or "YouTube thumbnail A/B testers." Mature niche = educated buyer = short sales cycle.

2. Ship a boring stack in a weekend. Docker Compose, one $12/mo droplet, Postgres, Stripe, Cloudflare in front. If you're spinning up managed services before you have 10 users, you're avoiding the product.

3. Write 40 programmatic SEO pages in month one. Each targets one long-tail buying query: "best [tool] for [use case]," "[competitor] alternative," "how to [task] in [platform]." These pages are your Google and your LLM-referral play. Expect 3–6 months to index.

4. Ship an MCP endpoint before your competitor does. Whatever your product does, expose it to agents. Add a /mcp route this week. In 2026 it's the only new distribution channel with zero saturation. Every agent user who tries you tells three others.

5. DM your first 10 paying customers personally. Ask what almost stopped them from buying. Ship the answer inside 72 hours. That's your v2. Do not build a roadmap.

Stack: DigitalOcean ($12/mo) + Docker Compose + Postgres + Stripe + Cloudflare + a weekend of MCP wiring. Setup: 3–5 days from empty repo to paid.