<p>You refresh the npm page once or twice like someone checking a barometer, close the tab, and try to get on with your day.</p><p>A week later, the first real numbers are in. And honestly? They're nicer than I expected.</p><h2>The numbers</h2><p>The three core libraries that make Elyra tick — <code>@elyracode/ai</code>, <code>@elyracode/tui</code>, and <code>@elyracode/agent-core</code> — are each pulling in around <strong>5,000 weekly downloads</strong>. The CLI itself, <code>@elyracode/coding-agent</code>, is sitting at <strong>3,600</strong>.</p><p>That's the kind of split that makes sense when you look at the architecture. The core libraries get pulled in transitively whenever someone installs the CLI, plus whenever any extension or downstream project depends on them. The CLI number is the one that maps most directly to <em>people who installed Elyra to actually use it</em>. So roughly 3,600 humans (and, fine, probably a few CI pipelines and curious bots) ran <code>npm install -g @elyracode/coding-agent</code> in week one.</p><h2>Putting it in perspective</h2><p>Let's be honest with ourselves for a second. OpenCode does 7.5 million monthly devs. Codex CLI has 75K GitHub stars. There are coding agents out there with download numbers that would make Elyra's first week look like a rounding error.</p><p>But that's not really the comparison that matters. Elyra is a brand-new, terminal-native coding agent in a space that already has loud, well-funded incumbents. The interesting question isn't "did we beat OpenCode" (we did not, and we will not be doing that next Tuesday either). The interesting question is: <strong>did anyone show up?</strong></p><p>The answer is: yes, a few thousand people did. They typed <code>npm install -g @elyracode/coding-agent</code> into their terminal. They set an API key. They ran <code>elyra</code> and watched a TUI flicker to life. Some of them probably hit a bug. Some of them probably bounced. But some of them stuck around, because that's how these things go.</p><h2>What 3,600 means in practice</h2><p>A nice mental model: imagine a small, packed bar. Maybe a beer hall. 3,600 is the kind of number where you could fit everyone who tried Elyra into a single venue if you really pushed it. That's not a faceless metric — that's a room full of developers who, for whatever reason, decided this week to give a new agent a shot instead of sticking with what they already had open.</p><p>That's flattering. It's also a lot of responsibility. Because every one of those installs is a developer extending you a small amount of trust: <em>I'll let your code run in my terminal, hold my API keys, and touch my files.</em> You don't get that for free, and you don't keep it without earning it.</p><h2>What I'm watching next</h2><p>A few things I'm curious about over the coming weeks:</p><ul><li><p><strong>Retention vs. drive-by installs.</strong> Weekly download counts can't really tell these apart. The shape of the curve over the next month will be more informative than the first-week peak.</p></li><li><p><strong>Extension downloads.</strong> Are people pulling in <code>@elyracode/stack-tall</code>, <code>@elyracode/flux-ui</code>, or any of the other extensions? That's a signal of people moving past "try it once" into "set it up for my actual stack."</p></li><li><p><strong>GitHub issues.</strong> Honestly the most useful metric of all. Bug reports mean engagement; thoughtful feature requests mean someone is thinking about how Elyra fits into their workflow.</p></li></ul><h2>A small thank-you</h2><p>If you're one of the 3,600 — thank you. Genuinely. You didn't have to try this. There are many shinier options, with bigger teams and slicker landing pages. The fact that you spent a few minutes of your one finite life pointing your terminal at a brand-new package made by a small team means something.</p><p>I'll keep shipping. You keep telling me what's broken. Deal?</p>