nextDash

Everyone's bookmarks end up the same way: a bar full of favicons with no labels, folders you never open, and a long tail of links that quietly stopped working years ago. The problem is not storage. A bookmark collection is write-only: you add to it constantly, you search it almost never, and nothing in the system ever tells you a link has gone.

Hosted tools solve the organising part, but they ask for an account and a copy of your links on somebody else's server. That trade never sat right for something as personal, and as boring, as a list of URLs.

Then the homelab joined in. Once the dashboard knew your links and could tell when they broke, it made sense for it to know the containers behind them too. Eight things nextDash does that a bookmark bar does not:

$ Keys: see it in action →

$ Find: see it in action →

More than 250 shortcuts across a dozen groups, but you need about ten of them to be faster than you were before. One rule holds the scheme together: every action on a bookmark is Shift plus a letter. Shift+E edits, Shift+V previews, Shift+L shares, Shift+S always opens config. Bare letters act on the first press. A short tour of seven steps shows the first ones (Shift+K plays it again), and a key legend under the grid keeps them in sight (Shift+Z hides it).

Three levels of structure, a handful of views

page          a tab across the top - "main", "work", "media". Keys 1-9.
 └─ category  a column heading - "vps", "tools", "reading". Collapsible, sortable.
     └─ bookmark   a row - name, URL, favicon, shortcut, tags, note, health status.
 └─ widget    a block among the categories, one or two columns wide.

The first ten keys

Shortcuts that open sites

Give a bookmark a shortcut - GITHU, JELLY, TOWER - and typing it on the dashboard opens the site. No search box, no Enter. The project's own note: "a shortcut that needs Enter to finish is not much of a shortcut." Collisions are real, so there are three modes under Config -> Behavior -> Search: open instantly (the default), open after a short pause, or require Enter.

Six ways to search, one bar

Filters are typed straight into the bar, with autocomplete after each prefix. A leading minus excludes instead of selects, so "tag:dev -status:pinned" means dev links I have not pinned. Your browser's address bar is a search box too: type the keyword, Tab, a term, Enter, and you land in your bookmarks with the results on screen.

> Try it: > to search, : for commands, ! for the cheat sheet. open the demo ↗

While search, the cheat sheet, quick-add or an inline edit is open, keyboard focus stays inside that overlay and the grid behind it is inert, not merely dimmed. That is the difference between a keyboard UI and a mouse UI with hotkeys bolted on.

$ Capture: see it in action →

$ Inbox: see it in action →

Saving a link should take almost no effort, or you will not do it. There are six ways to add links by hand, sources that bring links in on their own, and an inbox for everything that has not found its place yet.

Sources: links that arrive on their own

Config -> Data & backups -> Sources is for the places bookmarks keep arriving from, as opposed to the import buttons, which read a file once. A source remembers what it already brought in and previews what it would write before writing it. GitHub stars and Raindrop.io need a token; Hacker News, YouTube and Mastodon need nothing but a username or channel. A browser bookmark file imports too, and so do exports from Pocket, Pinboard, Raindrop, linkding, Shiori, Linkwarden and Karakeep, since they all use the same format.

Save something twice and nextDash asks instead of guessing. A URL already on the same page is refused; a copy on another page names the page and category it is already filed under and offers to open it, and "Save anyway" keeps the second copy.

Shift+J shows the focused bookmark as a QR code, so you can open it on your phone without typing the address, and Shift+L shares it.

The Inbox: triage, not a pile

Shift+I opens the holding area for links worth keeping before you know where they belong. A rail of filters - All, Unread, Snoozed, With note - narrows it, and you can sort newest first (the default), oldest first, by title or by site. A Tags column shows each link's own tags, and a dashed chip offers one more: click to add it, the cross to turn it down, or Add all for the whole list.

> Try it: Shift+I open the demo ↗

$ Health: see it in action →

$ Alerts: see it in action →

$ Statistics: see it in action →

This is the part no other bookmark manager has, and the reason nextDash exists. One view for the whole collection: filters for broken, stale, unused and duplicated, a side panel with Details, Health and Usage, and Work through to clear a pile one link at a time.

> Try it: Shift+H or Shift+U open the demo ↗

Three levels of checking

What it can detect

A self-hosted service bookmarked at its web interface would answer "not signed in" to an anonymous check. Store a sign-in once under Config -> Data & backups -> Sources -> Health sign-ins and point any number of bookmarks at it. The secret stays out of your backups unless you ask for it, and is dropped if the service redirects the check elsewhere.

Uptime that does not flatter itself

Uptime pools individual checks rather than averaging per-monitor percentages, and every percentage carries the number of checks behind it: 100% from three checks is a weaker claim than 100% from three hundred. Days you did not open the dashboard leave a gap rather than a straight line through them.

The whole collection at a glance

Collection health, reached with h from the Bookmarks view, is the wider picture: the score over time, what is wrong and of what kind, health per page, checking coverage, monitor uptime with the least reliable bookmark named, and certificates running out. Every number on it opens the matching filter. It has three tabs: Overview, Monitors (uptime per monitor, the least available and the outages as a 30-day lane each) and Trend (90 days, every monitor per day and uptime by weekday).

Getting told, and working the list

A downtime alert fires after a number of consecutive failures (default 3) and again on recovery. A container that stops or an Unraid disk that drops out reaches you through the same channels.

• An alert through ntfy carries Open link and Health buttons, so you can act on it without finding a laptop.
• Apprise reaches mail, Matrix, Signal and a hundred more. Everything about checking and telling lives in one section, Config -> Monitoring, with tabs for Checks, Notifications and Schedules (maintenance windows and quiet hours).
• A weekly digest sums up what needs you once a week: links that broke, certificates running out, container updates, and bookmarks nobody opens. A card on the dashboard, Your week in nextDash, opens it in a window with a button to the view where each part is dealt with.
• Tell it to leave one alone: press n or z on a row and that one condition is set aside while the bookmark keeps being checked. An Ignored filter lists everything you have put aside.
• Ten links, two minutes: a corner card names what is waiting and runs Work through over the worst ten.
• Repair, not just diagnosis: follow redirects in one action after a domain move, or point a dead link at its last Web Archive capture.

$ Containers: see it in action →

$ Status page: see it in action →

$ App icons: see it in action →

A bookmark dashboard that knows your Docker host too: every container on the machine, grouped by compose project, matched to the bookmark you already have for its web UI.

running unhealthy update available stopped

> Try it: Shift+Y open the demo ↗

Health follows the bookmark

A web UI's bookmark carries its own health mark on the container row - green where the monitor finds it up, red when it does not. The side panel says why: failed healthchecks, a crash, a restart loop since the last update - the same diagnosis Bookmarks would give you, without leaving the Containers view.

Updates, with a way back out

Pick containers that update themselves in a window you choose. Each one is watched for five minutes after; turn unhealthy in that window and nextDash puts it back on its old image and skips that version, with the reason in its Timeline and a notice sent to your phone.

An update or rollback shows its steps while it runs - check, pull, compare, tag, stop, rename, create, connect, start and remove - with a log line for each pulled layer. Update a selection and nextDash takes only the containers that have an update, and says why one was left out. Services get their real app icons from two open icon sets, matched by image and host, with a picker behind the pencil.

Logs follow a container live in their own window. The Disk tab shows what images, volumes and build cache take up and clears it after naming what goes. A container that stops, keeps restarting or runs hot reaches you through the same notices and alert channels as a downed bookmark - one place to learn something on your network needs attention, not two.

$ Unraid: see it in action →

For an Unraid box, nextDash talks to Unraid's own API - a Viewer-role key is all it asks for - and turns the array, the parity check, shares, VMs, the UPS and notifications into seven read-only widgets next to the bookmarks you already have.

Read-only by design: a Viewer key can show you a parity check is running, it cannot start or cancel one. A disk dropping out of the array, a parity error, or the UPS switching to battery reaches you through the same alert channels as a downed bookmark or a crashed container.

$ Widgets: see it in action →

$ Custom widgets: see it in action →

$ Moving blocks: see it in action →

$ Notes and QR codes: see it in action →

A widget sits among your categories, is dragged into place like one, and can be one or two columns wide. Add, name and arrange them under Config -> Widgets, which has three tabs - Your widgets, Catalogue and Custom widget - and an Add a widget picker you can search. More than thirty built-in tiles, grouped by the question each one answers, and new kinds carry a New badge.

Custom: the escape hatch

For a service that is not in the list, point a Custom widget at an address that answers with JSON, name the fields, and it draws them. Over forty self-hosted services come pre-filled, from Pi-hole and AdGuard Home to Proxmox, Grafana and Nextcloud. It can use the same stored sign-in as a health check, so no key ever reaches your browser.

$ Looks: see it in action →

$ Theme studio: see it in action →

This is a page you look at every day. nextDash ships over 300 themes - 160 families with a dark and a light half - and each has a character of its own, not just a different palette. Twelve characters decide what kind of surface a theme draws: lacquer, glass, frost, aurora, neon, velvet, enamel, brushed, carbon, paper, terminal, ink. Glass is something you look through; velvet keeps its light in the shadow beneath; neon lets the accent leave the surface as a halo.

The same dashboard, four looks. Same data, same layout - only the studio settings change.

glass aurora neon terminal

> Try it: Shift+A open the demo ↗

The studio opens beside your dashboard and changes the real page as you go: the theme, a couple of dozen backdrops drawn in its colours, depth, glow and card glass, the category headers, the type and the contrast - or one of a dozen ready-made looks that set them all at once. Compare shows what you had, a dice icon tries something at random, and nothing is saved until Apply.

Twelve themes, each in light and dark

The page is the same every time. The theme is not, and neither are the settings that go with it: look, backdrop, glass, depth, glow, headers, type and density. Below is a gallery: click through the twelve themes and switch each one between light and dark.

Mode

☾ dark ☀ light

← prev 1 / 12 next →

> Pick a theme below, flip between dark and light, or use the arrow keys. Click the big picture to enlarge it.

lacquer glass frost aurora neon velvet enamel brushed carbon paper terminal ink

Config, organised and not overwhelming

Appearance has nine tabs in three groups, and its panels fold up to a line showing what they are set to. Every section is deep-linkable and opens on cards for the few settings a group leads with, rather than a wall of every option. An At a glance panel shows how many settings differ from default, an Only changed filter sits above every tab, and Ctrl/Cmd+Shift+K finds a setting by name or by its current value.

$ Copies: see it in action →

$ Collections: see it in action →

Three things that make the library keep itself in order: copies that outlast the original, collections that reorganise around what you actually do, and hooks so other tools can listen and ask.

Keeping a copy of a page

Collections, tags and Fresh

Today (your keyword sets), Recently opened, Most used, Stale, Fresh (bookmarks whose site published something new), one collection per tag, and your own rules on category, tag, page, status, pinned, untagged, or days since last opened. Fresh reads the head of every saved page, notes any RSS or Atom feed advertised there, and polls it hourly. It stores no articles, only how many entries are newer than your last visit - it is not a feed reader.

Shift+T opens a quick-tag popover. / opens the tag cloud: three sizes by use, each with its count, and you type to find a tag. Picking one opens the tag view, with every bookmark that carries it from every page, related tags, bulk actions, and rename, merge or remove for the tag itself; t opens it for a bookmark's own tag. nextDash also suggests tags, from a catalogue of some 51,000 sites and, for the rest, Wikidata: a dashed chip in the row, one click to add it, or Add all. Preview cards show who published a page and when, and for video providers they include a player you can start without leaving the dashboard.

Telling other programs, and letting them ask

Config -> Data & backups -> Webhooks pushes six events to a receiver: a bookmark added, changed or removed, a monitored bookmark going down or coming back, and a certificate about to expire. They are signed with the Standard Webhooks scheme, so a receiver that already verifies those needs no special case; a failed delivery is retried twice, and endpoints are checked at save time and again at delivery.

POST /your-endpoint
content-type: application/json
webhook-id: msg_2b7f...
webhook-timestamp: 1756253400
webhook-signature: v1,K5s0...
{"type":"bookmark.added","timestamp":"2026-08-27T09:30:00Z","data":{...}}

One MCP endpoint at POST /mcp lets an assistant search the collection, look one bookmark up, list the tags in use and add a bookmark. It is off until you switch it on, because it answers questions about every bookmark in the install - and adding a bookmark through it goes through the same handler the dashboard posts to, so duplicate checks and webhooks still apply.

$ Outgoing and trust: see it in action →

$ Data: see it in action →

Your bookmarks are a map of what you care about. That map should live on your own disk - and the way a project handles that question tells you most of what you need to know about it.

• The API answers this dashboard, not every page on the internet: cross-origin requests are refused unless you allow-list them.
• A stored sign-in stays on the service it was stored for, and a custom widget's address and key stay on the server.
• A webhook cannot be talked into delivering somewhere else, and outbound fetches are defended against SSRF at dial time, including DNS rebinding.
• The data directory is not browsable: it publishes icons and an uploaded favicon or font, and nothing else.
• A few requests go out by default, each with its own switch under Behavior -> Privacy & sync: an install count once a day (a random id and the version number), a daily check on GitHub for a newer release, and the nextdash.cc news feed. None of them carries bookmarks, settings or addresses. DISABLE_TELEMETRY, DISABLE_UPDATE_CHECK and DISABLE_NEWS_FEED turn them off for the whole server.
• Analytics are opt-in and off by default. When off, the tracker script is not even emitted into the page. When on, no bookmark names, URLs or search queries are sent, and every number is rounded into a band. DISABLE_TELEMETRY=true removes the choice server-wide.

Your data is a folder

There is no database. Your entire installation is a folder of JSON files you can read in a text editor, copy to a USB stick, or put in git.

data/
├── bookmarks-1.json    one file per page: the page, its categories, its bookmarks
├── settings.json       every setting, saved searches, collections, theme choice
├── colors.json         your custom themes
├── inbox.json          captured links waiting for triage
├── trash.json          deleted bookmarks, kept 30 days
├── health-history.json uptime samples - 30 days raw, 90 days of daily rollups
├── webhooks.json       endpoint URLs and signing keys, mode 0600
├── icons/              fetched and uploaded favicons, content-named
├── local-copies/       saved page snapshots
└── auto-backups/       rotating ZIPs, 3 by default (1-50)

A backup carries your whole data directory, so a restore does not lose your history. Two switches let you leave out the two largest or most sensitive things: local copies, and stored tokens. Exports exist for every part - bookmarks as CSV or browser-compatible HTML, inbox as CSV or JSON, health state and raw checks as CSV - and deleted bookmarks stay recoverable for thirty days in the Trash.

Is this for you?

$ Install: see it in action →

$ Page templates: see it in action →

I make nextDash for myself first: it is the start page on my own screen, and the one place I look to see whether my homelab is fine. If that sounds like something you would use, there are three ways in.