I Launched to Nobody
Source: Dev.to
The Idea
Every time I start a new AI coding session, I re‑explain the same things: the project architecture, the coding standards, which patterns to follow, which to avoid. It’s the same context, over and over, across every tool—Claude Code, Cursor, Copilot. They all start from zero.
So I built ContextPin, a local desktop app where you write your project context once, in a nice editor, and every AI tool reads it automatically through MCP (Model Context Protocol). No copy‑pasting. No “here’s my project structure” at the top of every chat.
Three Months of Building
First commit: November 1 2025
Beta shipped: February 3 2026 (three months, solo)
The app grew into something I’m genuinely proud of:
- A rich Notion‑like editor
- Workspaces, drag‑and‑drop context trees
- Full‑text search, cross‑references, mermaid diagrams
- Version history, encrypted local storage
- A built‑in MCP server that any compatible AI tool can connect to
January alone saw ~90 commits in 24 days—mostly evenings and weekends. The editor went through dozens of iterations (list parsing, paste handling, code blocks, tab caching). Getting a rich‑text editor to feel right is one of those problems that looks simple until you actually try it.
I also had to get the app signed and notarized by Apple; without that, macOS shows a scary “unidentified developer” warning—a death sentence for an unknown app. The Apple approval process took about eight days of waiting.
In parallel I built:
- The licensing API
- The marketing website (Astro + Tailwind on Cloudflare)
- Payment integration and analytics
All the usual startup infrastructure.
February 3: Tagged v0.1.0-beta.30 as the launch version and pushed the website live. I shared it on Hacker News, posted on Reddit, and spread the word wherever I could—then waited.
The Numbers
Two weeks after launch I finally looked at the funnel with honest eyes.
| Metric | Count |
|---|---|
| Pageviews | 185 (real) |
| Downloads | 22 (all me testing) |
| Trials started | 8 (all me testing) |
| Checkouts | 5 (all me testing) |
| Subscriptions | 1 (me) |
185 real visits. Zero real downloads.
I had polluted my own analytics by testing every step of the funnel myself. The numbers that looked like a conversion problem were actually just… me, talking to myself.
The real story was much simpler: 185 people visited the site. Not a single one downloaded the app.
What Went Wrong
I went back to the landing page and tried to see it through a stranger’s eyes. It wasn’t great.
-
Headline: “Local‑First Context Editor for Claude Code.”
If you already know what ContextPin does, that makes sense. For a random developer landing on the page, it means nothing. “Context Editor” describes the mechanism, not the problem it solves. -
Zero social proof. No testimonials, user count, GitHub stars, nothing. An unknown developer asking you to download and install a desktop app with no trust signals? Hard no for most people.
-
7‑day free trial buried in the FAQ. The main CTA just said “Download.” No indication of what happens next, whether it costs money, or how to try it risk‑free.
-
Pricing hidden on a separate page. $9 / month is actually very reasonable for a developer tool, but visitors who can’t find the price assume it’s expensive.
-
Best demo (interactive compare view showing the editor vs. markdown output) was too far down the page. Most people probably never scrolled to it.
My Fixes
- New headline: “Stop Re‑Explaining Your Codebase to AI.”
- Restructured the page to lead with the interactive demo.
- Made the trial front‑and‑center.
- Put pricing on the landing page.
- Rewrote every section to focus on the pain, not the features.
The Bigger Decision
The landing page wasn’t the only problem. I also realized the 7‑day trial model itself was broken for what I’m building.
Developers typically:
- Download a tool, play with it for ~20 minutes.
- Get pulled into real work and forget about it.
- A week later the trial expires—no habit formed, no moment where the tool becomes part of the workflow.
MCP is a new concept; it takes time for developers to internalize what “persistent AI context” actually changes. A countdown timer kills that exploration.
New Model: Freemium
- Free forever with limits: 1 workspace, ~30 contexts, no version history, no live contexts.
- Full editing, full MCP access, full editor quality.
- The free tier should be genuinely useful for one small project.
When someone needs a second workspace or their context library outgrows 30 documents, upgrading to $9 / month becomes a no‑brainer.
Inspiration: TablePlus—useful for free, mildly annoying at scale, pay to remove the friction.
Is the free tier too generous? Maybe. But the risk of having zero users is infinitely worse than the risk of under‑monetizing. I can tighten limits later when I have real usage data.
What’s Next
- Ship the landing‑page changes.
- Build the freemium model.
- Drive traffic to the site (185 visits in 30 days means almost nobody knows this exists).
I’ll keep posting updates as I figure it out. If you’re building something and struggling with the same “launched to nobody” problem, feel free to reach out.
I’d love to hear about it.
You can check out ContextPin at contextpin.com.