Last month our team lead forwarded a Notion invoice and asked a boring question: which of these docs do we actually still use?
Nobody knew. That's the thing about a wiki that lives in someone else's cloud — you never quite know how much of your institutional memory is sitting there, or what happens to it the day the seat count gets cut. So I spent a weekend going through the open-source Notion alternatives, and I did it the unglamorous way: a spreadsheet, three columns, honest scores.
Here's what I learned, and where you can steal the numbers.
The three columns that actually decided it
1. Data ownership. Not "is it open source" — that's table stakes. The real question is: if the company behind it disappears tomorrow, can I still open my files? A self-hostable server with a SQLite dump that I can read scores high. An Electron app that phones home to a proprietary sync layer scores low, no matter how pretty the UI.
2. Offline reality. Every app says "offline support". Most mean "we cache the last page you opened". I tested by killing the network mid-edit and reopening the app cold. Apps that came back with a full tree and a working editor went up a rank. Apps that showed a spinner for 40 seconds slid down. One app displayed my pages but silently refused to save — that's worse than failing, and it cost it the trial.
3. The migration cliff. This is the one people skip. Ask: what's the escape hatch? Not import from Notion (everyone does that). I mean export out — to Markdown, to plain folders, to something a future tool can read. If the export produces a mangled folder tree with UUID filenames, the answer is no.
The honest shape of this: I didn't want a better Notion. I wanted a system where the exit cost was near zero. Once you score for exit cost instead of features, the shortlist collapses fast.
What the numbers look like
I ended up shortlisting three, and the thing that surprised me was how different their answers were.
- AppFlowy — roughly 79k stars on GitHub, AGPL-3.0. Flutter, desktop-first, and it genuinely targets the Notion mental model (databases, boards, wikis in blocks). Export to Markdown works. Weak point: sync still wants their cloud unless you're prepared to run the stack yourself.
- AFFiNE — MPL-2.0, the docs+whiteboard hybrid. If your notes are 60% diagrams, this is the only one that stops feeling like a compromise.
- SiYuan (思源笔记) — AGPL-3.0, local-first to its bones, encrypted-by-default. Best privacy story of the three, steepest learning curve, and the least "Notion-shaped" — which for some people is the point.
All three live on GitHub with real release pages, so you also get the boring-but-important part: you can see what changed last week, instead of trusting a changelog page nobody maintains.
The part where I stopped reading blog posts
The failure mode of this kind of research is spending two hours on listicles that were all written from the same press release. What actually helped was a page structured around the decision instead of the product — the kind that lists each replacement with its license, whether self-hosting is real or nominal, and what the migration path is.
I used GitHubPlay's Notion alternatives page for the sweep — it indexes them with OSI licenses and self-hosting verification, and each entry links straight to the official GitHub Releases rather than a mirror, which matters when you're about to install something on a work machine. For AppFlowy specifically, the canonical repo page has star counts and downloads synced from the GitHub API, so you're looking at live numbers instead of a screenshot from 2023.
Caveat I'd want stated if I were reading this: these numbers move daily. Treat any star count in a blog post, including this one, as a timestamp, not a fact.
What I'd tell you to do instead of picking from this list
Don't pick from this list. Do this:
- Pick your current worst moment. For us it was "someone shares a doc, the offline copy is stale, and they edit the wrong version." That's the thing to test.
- Install two, not five. Trial period should be measured in hours, not weeks.
- Kill your network and try to work. Ten minutes. That's your real answer.
- Export everything once, before you migrate. If the export is garbage, you learned that cheaply.
- Only then decide whether "close enough to Notion" even matters to you. Turns out, for us, it didn't.
We ended up keeping a self-hosted instance and a Markdown folder that gets committed to a private repo every night. Boring architecture, near-zero exit cost. The invoice at the top of this post is now a cron job that costs nothing.
If you're doing this comparison yourself, the alternatives index groups them by the SaaS tool you're leaving, which is a faster way in than searching "open source Notion alternative" and drowning in SEO farming.