Private preview links
Pagoti lets you share markdown privately without requiring anyone to log in. The concept is straightforward: generate a private preview link, share it, and revoke or refresh it whenever needed.
The problem...
A lot of Pagoti content starts private. Until now, sharing usually meant screenshots, PDFs, or manual exports — unnecessary friction.
Private preview links provide a clean way to show work-in-progress content to a client, teammate, or friend without making anything public or asking them to create an account.
How the feature works
Pagoti generates a private preview link tied to a project. The link includes a token so it can be viewed without authentication.
The link can be refreshed at any time, which immediately invalidates any previously shared versions, or removed entirely to disable access.
All of this lives behind a dedicated modal in the dashboard. It shows the active link, offers a one-click open, and protects refresh and remove actions with confirmation prompts so nobody accidentally locks someone out.
When you'd actually use this
This feature exists because of very ordinary situations:
- Sharing a draft blog post with someone for feedback without publishing it.
- Sending a spec or proposal to a client without managing access or accounts.
- Iterating on documentation internally where someone needs to read, not edit.
- Sense-checking an idea before deciding whether it should ever be public.
In all of those cases, the goal is the same: send a link, let them read, stay in control.
What changed in the dashboard
- A new "Private preview" action now lives in project settings.
- Refresh and remove actions are protected with confirmation modals to prevent accidents.
Building with AI, candidly
This feature sounded simple on paper: share markdown privately using token-based links, no login required. In practice, it took a bit more work than expected.
Planning through Claude in detail didn't land first time. Or the next. The early attempts were close enough to feel promising, but off in ways that mattered.
The solution came from restarting and trickle-prompting through the problem, tightening constraints and correcting assumptions along the way.
That's a common experience with AI on anything non-trivial. One big prompt rarely gets you there. Incremental steering does. A clear idea of what needs to be built is still required, along with the judgement to spot when something looks right but isn't.
AI sped things up overall, but it didn't remove the need for judgement. If anything, it made that part more visible.