#1 on npm. One real user. Both true at the same time.
We're the most-installed MCP memory server on npm. We have exactly one real user. We're going to tell you why both of those numbers are correct, why neither one is failure, and what we are actually trying to do here.
The absurd part
On May 23, 2026, @cachly-dev/mcp-server crossed 472 weekly installson npm. The next four memory-focused MCP servers added together don't reach that number. We are, by the cleanest metric available, the #1 MCP memory server in the world.
On the same May 23, 2026, the count of humans who signed up on cachly.dev without knowing us personally and used the product is: one.
One. Singular. A human. From a country we've never been to. We've been refreshing the admin dashboard like it's a slot machine.
Both numbers are real. We checked twice.
The funny part
You can read those two numbers and conclude: they're failing. You'd be in good company. We thought the same thing for about four hours on Tuesday.
Then we noticed something. The 472 weekly npm installs are happening because thousands of developers are quietly using cachly through their AI editor — Claude Code, Cursor, Copilot, Windsurf — without ever visiting our website. They get the brain. They never meet us. They're using cachly the way you use DNS: silently, every day, and you don't think about it.
And the single signup? Not a defeat. A signal. The first person who not only used the product but cared enough to put their name on it. That's the first stone in the wall.
The serious part
We're writing this down because we keep meeting founders who interpret “most installs in our category but no signups” as a verdict on whether they should keep going.
It is not a verdict. It is a piece of information.
The two ways to read it:
- Reading A (defeat): “Lots of people install it. None of them love it enough to convert. Therefore the product is wrong. Pack up.”
- Reading B (path): “Lots of people install it. We haven't given them a reason to come meet us yet. The product is one half. The relationship is the other half. Build the second half.”
Both readings fit the data. Edison didn't fail 999 times before the lightbulb; he found 999 materials that don't conduct electricity well enough to glow without melting. Same dataset, different sentence. The sentence you choose decides what you do next.
We choose Reading B. Not because it feels nicer — though it does — but because Reading A doesn't explain the 11,161 monthly downloads. Reading A would have us shipping something nobody wants. Reading B has us shipping something a lot of people want, quietly, and we haven't finished saying hello yet.
What we're actually trying to do
We want to take the cost of forgetting away from every human who has ever earned a piece of knowledge the hard way.
That sentence is the whole thing. Read it again if you want.
Right now we're doing it for developers, because developers are the people we are, and the people whose forgetting we can measure most precisely. The 3-hour bug a developer fixed on a Tuesday in March is the same shape as the work-around your parent figured out for their tax software, the deployment checklist your colleague keeps in a notebook nobody else can read, the medical answer a nurse gave a patient at 2am that three other patients needed the same week.
All of these are humans paying twice — first to learn, then again to re-learn because the first learning evaporated. That's the tax of forgetting. It compounds. It is one of the largest, most invisible costs in human work, and it's completely solvable with software that doesn't forget on your behalf.
So that's the bet. We build the memory layer for AI assistants first because the AI assistants are the people (well, the things) most obviously suffering from forgetting right now. The same primitives — typed cause-and-effect graph, ambient learning, distilled crystallized briefings — generalize to humans, teams, and entire organizations.
The mantra
We say this one to each other on bad days. It's also above the door of the GitHub Actions workflow file.
Once learned, always remembered.
For every human. By default. Forever.
That's it. We don't have a clever follow-on. The three lines do the work.
What the bug report doesn't say
You should also know: we have a lot of bugs. We have a screen on the dashboard right now where the empty state says “Loading...” for three seconds before it says “No data.” We have an iOS build that we couldn't finish this week because Apple ID password reset is its own journey of personal growth. We have one user-reported case where the VS Code extension doesn't reconnect after a laptop wakes from sleep. Sometimes our newsletter signup flashes “something went wrong” for half a second before succeeding.
We don't pretend these aren't there. Each of them is a way that, today, doesn't work. None of them is a verdict on whether the thing is worth building.
We just write the bug down, store it in our own Brain (cachly eats its own dog food and the dog food is memory), fix what we can on Friday, and ship the rest next Tuesday.
The deal we're making with you
If you read this far, here is what we owe you in exchange:
- We will keep shipping in public. Every release, every mistake, every quiet metric. Nothing in the dark.
- The Free tier stays Free forever. 25 MB of brain. All 84 tools. No credit card. No expiry. No bait. If your AI ever stops remembering, it won't be because we made you pay.
- Your data stays on German servers. GDPR by default, not as an upgrade. We will never sell, train on, or otherwise monetize the things you tell your AI in private.
- When we fail at any of the above, we will tell you. In a post just like this one.
And if you want to be user number two
You don't need to do anything dramatic. One command.
npx @cachly-dev/initThat'll set up your AI editor with permanent memory in about thirty seconds. The next time your AI starts a session, it'll already know your stack, your last deploy, and the bug you fixed in March.
If you sign up at cachly.dev afterwards, we will notice. There is a tiny bell on our admin dashboard. It will ring. Someone will yell across the room. That's the entire onboarding ceremony for now. We'll upgrade it once there are enough of you that the yelling stops being practical.
cachly is a persistent AI Brain for developers — memory shared across Claude Code, Cursor, GitHub Copilot & Windsurf simultaneously. Auto-detects every editor. Bootstraps from your git history. 89 MCP tools. Free tier, EU servers, no credit card.
Your AI is forgetting everything right now.
Every session starts blank. Every bug re-discovered. Every deploy procedure re-explained. cachly fixes that in 30 seconds — your AI remembers every lesson, every fix, every teammate's hard-won knowledge. Forever.
Once learned, always remembered. For every human. By default. Forever.