Why this was written
We build an Obsidian plugin, so we wanted to know what actually separates the few Obsidian plugins that millions of people install from the thousands that never pass a few hundred downloads. Opinions on this are easy to find. Numbers are not, even though Obsidian publishes them: the download count of every community plugin is in a public file on GitHub.
So we took that file on 8 October 2026, compared it with snapshots from a year and two years earlier, and read the top of the list plugin by plugin. Claude drafted the analysis from the public data in one sitting. The page lists its sources at the end and marks which parts are interpretation rather than measurement.
The result is a one-page report, and it is a working example of the format this site is about. It is a single Markdown note. It renders as the page above, with stat tiles, a table, cards and a numbered list, and the same note opens in Obsidian.
How to read it
- Downloads are not users. Obsidian counts every install and every update. A plugin that ships often gets counted more. Claudian's 90 releases in its first year inflate its number, and the page says so.
- "Last 12 months" is the gain between the October 2025 snapshot and October 2026, which shows momentum better than the lifetime total.
- Part 3 is interpretation. The seven patterns are a reading of which plugins lead and what they have in common. They are not a measured result, and the page labels them that way.
- Revenue figures are third-party. Copilot's revenue comes from TrustMRR's read of its Stripe account, not an audit.
- Marketing channels are a gap. We could not verify which channels (YouTube, Reddit, newsletters) drove the winners, so the page leaves that as a hypothesis.
Key takeaways
- It is a power law. The top 1% of plugins take about 65% of all downloads, and the median plugin has about 830.
- The catalog more than tripled in a year. 8,566 plugins against 2,635 a year earlier, after Obsidian moved first-submission review to automated scans in May 2026. Listing is now easy, and gives almost nothing by itself.
- Winners fill a hole in core. Templater, Calendar, Tasks, Kanban and Advanced Tables each fix something a new user runs into in the first week (Advanced Tables, for example, makes Obsidian tables easy to type).
- Winners keep data as plain Markdown. Uninstall them and your notes are still notes. That is the promise Obsidian users chose it for.
- Each AI wave makes a new leader quickly. Claudian, which brings Claude Code and Codex into Obsidian, went from launch in December 2025 to about 2.3 million downloads in under a year.
- Core can absorb a category. Canvas and Bases went straight at Excalidraw and Dataview. Their downloads show no visible dent yet, but a late entrant in a space core covers has little room.
- Donations do not pay. Templater, one of the three most downloaded plugins, had six active sponsors when its author raised it. The model that pays in the data is a free local plugin plus a paid hosted service.
If you are building a plugin
Read as advice, the page comes down to a few questions worth asking before you write code:
- What hole in core does it fill? If a new user would not miss it in their first week, it will be hard to explain in one line, and the directory's search only matches name, author and description.
- Does the data stay plain Markdown? If uninstalling your plugin leaves notes broken or unreadable, users will hesitate, and say so.
- Is core likely to absorb it? Canvas and Bases show that Obsidian builds the most-requested features itself. Building next to core is safer than building where core is heading.
- Where will users hear about it? With hundreds of new plugins a week, being listed is not distribution. The research could not measure which outside channels work, so treat that as your own experiment.
- How will it pay? If it needs to, plan for a free local part and a paid hosted part from the start; Obsidian has no payment system, and donations rarely cover the work.
What we took from it
The patterns shaped myone.page, the plugin this site belongs to. It fills a hole: Obsidian has no way to share one note with a few people and get their edits back (Obsidian Publish shares a whole site, read-only). The page stays plain Markdown, with a few callouts and an optional comment line, so it reads as a note without the plugin. And the free local part (every block, theme and the HTML export) is separate from the paid hosted part (unlimited shared pages), which is the model the data says works. Whether it wins is a question for next year's snapshot. More on who builds myone.page and how.
Write your own research page
The same approach works for any market scan, competitor review or internal analysis: collect the numbers, let an AI draft the page in a fixed format, then check it. See how to write a one-pager with ChatGPT or Claude, or the one-page report template for the structure.