Collaboration & Sharing
Work together in real time — share gardens, see live cursors, and manage roles.
Sharing a garden
Click the share button in the top bar to invite someone by email. Only a garden's owner can share it — if you were invited to someone else's garden, the share button tells you so rather than letting you invite others. New invites start as Viewer — they can browse the garden but not change it. Use the role dropdown next to any member to switch them between Viewer and Editor at any time; the change applies immediately. Collaborators see each other's names and pictures; email addresses and pending invites are visible only to the garden's owner. You can also share a single node: drill into it first, then open Share — the invite covers just that node's canvas, and it opens for the guest as its own garden.
Sharing a link with anyone
Open Share and choose “Create share link” to publish the canvas you’re on as a page anyone can open — no account, no sign-in. What gets published is a snapshot: a copy of the canvas as it looks at that moment. Editing the garden afterwards changes nothing for people holding the link until you press “Update snapshot”, so you can keep working without wondering who is watching. The same one-layer rule as invites applies: the nodes on that canvas are published, and anything nested inside them is not. Pictures come along: an image in a thought’s description shows on the published page, an Image node shows its picture rather than the address it came from, and a YouTube node gets a real player — the thumbnail loads first and nothing is requested from YouTube until a reader presses play. A Link node arrives as a real preview card too — the page’s title, its description and its thumbnail, looked up once when you publish and stored alongside the snapshot, so nobody reading your garden fetches the PAGE itself — only the small picture or icon the card shows, and only from the site the link already points at. Every web address on a published page is clickable now, whether or not a preview was found for it. Some sites give out nothing about themselves to anything that is not a browser — recipe and news sites especially — and for those the card shows the site’s own icon and its address instead, the same as it does in your garden. A published card shows its typed properties the way your own card does — except the ones marked hidden, which are never copied into the link at all, so a hidden value is absent rather than merely unshown. Thoughts the AI suggested and you haven’t accepted stay behind too. Drill into a node first if you want to publish just that node’s canvas. “Revoke” turns the link off immediately and for good — a revoked address never works again, even if you publish the same canvas later.
Finding your way around a shared garden
A published garden opens fitted to the whole of it, which on a big garden means a lot of names too small to read. If you have named places in it — regions, or viewpoints you saved with the bookmark button — they are now listed for your readers as the garden’s contents. On a wide screen the list sits open at the top-left corner of the canvas; on a phone it folds into a single button in the same corner. Pressing a name flies the canvas to that place, so a visitor can tour a garden in the order you laid it out instead of panning around looking for the good parts. Regions and saved viewpoints appear together, each with its own small icon, in the order you made them — so the way you name and arrange things is the way a reader meets them. A garden with nothing named in it shows no list at all. Saved viewpoints travel as a framed area rather than as an exact camera position, so a reader on a phone lands on what you were looking at rather than on the corner of it.
Embedding a published garden
Alongside the link, “Copy embed code” gives you an <iframe> you can paste into a Miro board, a Notion page, a blog post or any site that accepts HTML. It shows the same read-only canvas without the surrounding page, and readers can pan and zoom it. Anyone reading a published garden can take that snippet too — “Copy embed code” sits in the page’s own menu next to “Copy to my garden”, so someone who wants to write about a garden you published can embed it without asking you for the code. It is the same snippet either way, pointing at the same address. The embedded view is the only page that may be framed by another site; every other page of the app refuses, and the embed is left out of search results so it never competes with the page it is part of. Published links also unfurl properly when pasted into social media and chat apps, with the garden’s title and how many thoughts it holds. The canvas keeps its own background too — the ring grid, and the soft colour glow around each type’s cluster of thoughts, captured as you had them when you published. A published page opens in whichever colour scheme the reader last chose, and the sun/moon button at the corner of the canvas lets them switch — the same control as the rest of the site. Readers can explore it, not just look at it: hovering or clicking a thought highlights it, and a [[wikilink]] whose target is one of the published thoughts is clickable — it pans the canvas to it, the same way following a connection does. A wikilink pointing at something you didn’t publish stays plain text, because there is nowhere for it to go. The same holds after a download: importing the JSON turns those links into real wikilinks between the imported copies, and the ones with no target arrive as plain words rather than as links that do nothing. Zooming works the way it does in your own garden — pull back and the cards shrink to compact cards, then to titled pills, then to labelled dots and finally to bare coloured specks, so a large published garden reads as a map rather than a wall of text. “F” fits everything on screen, or frames just the selected thought, and panning stops before the last thought would leave the view, so a reader can’t get lost in empty space. Pointing at a connection brings its label up to readable size whatever the zoom, and puts a small arrow at each end that takes you to the thought on the other side — they stay put while you reach for them. The page names you as the publisher — the display name from the account you signed in with sits under the garden’s title, next to how many thoughts it holds and when the snapshot was taken. It is captured when you publish, so a link you shared before you had a name keeps saying nothing about who made it until you press “Update snapshot” — which is also how a link already out in the world never gains a byline you didn’t choose to give it. If your account has no display name — or only an email address where a name would go — the page simply says nothing about who made it; your email is never shown. The download button saves the whole published garden as a JSON file in Neural Garden’s own format, so it can be imported straight into a garden of your own — thoughts, their text, where they sat, what they connect to, and the properties shown on each card. Types travel by their slug, so a reader who already has a “recipe” type gets the rows on their own version of it, and a reader who doesn’t is offered the chance to create it during the import. Regions don’t survive that trip, and neither do properties the type never gave a slug. On a wide screen the copy, embed and download buttons sit labelled side by side in the top-right corner; on a phone they fold into a single menu so the title has the room, and the button shows a tick for a moment after a download saves or an embed code is copied — both leave nothing on the page to say they worked, so the tick is how you know. If the tick does not appear, nothing was copied. Nothing a reader does changes what you published.
Taking a published garden home
“Copy to my garden” on a published page makes the whole thing yours — a real, independent copy that lands as one new node on your home canvas, with everything inside it. You arrive on your own canvas, the view pulls back to the new node and it flashes briefly so you can see where it went. If you don’t have an account yet, pressing it takes you through signing up first and the copy runs the moment you land; the link you clicked is remembered across the whole trip, including a sign-in link opened from your email in a new tab. What you get is exactly what you were looking at: the thoughts, their text and properties, where they sat, what they connect to, and the coloured regions grouping them. Types are matched to your own by their slug, so your “recipe” type receives the rows — and a type you don’t have is added to your Types garden for you, keeping the colour and icon the publisher gave it. Nothing of yours is ever changed to match theirs: if your “recipe” has different fields, yours stays exactly as it is. A [[wikilink]] between two published thoughts becomes a real link between your two copies; one pointing at something that wasn’t published arrives as plain words, because there is nothing for it to point at. If the name clashes with a garden you already have, the copy comes across numbered — “Pal’s Recipes (1)” — rather than being refused.
Listing a garden in the gallery
A share link reaches the people you give it to. The gallery — the Gallery page on the website, linked from the home page and from the bottom of every published garden — reaches everyone else. Once a canvas has a share link, the Share panel gains a “Show in the gallery” switch beneath it. Turn it on and the garden appears in the gallery for anyone to find, open and copy; turn it off and it disappears from the gallery while the link itself keeps working. Before or after switching it on, pick a few of the ready-made categories (Research, Writing, Product, Personal and so on) and add tags of your own — type one and press Enter — then press “Save tags”. Tags are what the gallery’s browse bar and search box match on, alongside the garden’s title and description, so a garden about sourdough is easiest to find when it says so. Up to eight tags; they are tidied into lowercase labels automatically, so “Second Brain” and “second-brain” are the same tag. What the gallery shows is what the link shows: the snapshot you published, with its title, your display name if you have one, and how many thoughts it holds. The picture on the card is drawn for you from the snapshot — every thought becomes a small card in its type’s colour, with its connections and regions — and it redraws when you press “Update snapshot”, so there is nothing to upload.
What a guest can reach
Access is per-canvas and explicit. Sharing a garden shares the nodes sitting on its canvas — not what's nested inside them. Guests don't see a node's nested items or backlinks, can't open it as a garden, and won't find its contents in search or the [[ link picker, until you share that node with them too. To open up more, drill into the node and invite them from there.
Contributing as a guest
Editors can create new thoughts on the shared canvas. Anything a guest creates is automatically theirs to work in — they can open it, nest inside it, edit and delete it — while the garden's owner remains the owner of everything in the garden. That also means contributions are permanent gifts: once your thought is in someone else's garden, moving it back out is not possible.
Your gardens stay yours
Each person's gardens are their own world. Working inside someone else's garden, the [[ link picker and search suggestions offer only that owner's content you can reach — your own notes stay out of their garden's pickers, and theirs stay out of yours. Existing nodes can't be moved or linked across two people's gardens; if you want your note in a garden you were invited to, create it there (or copy the content over). The global search (Cmd/Ctrl+K) still finds everything you can access, for jumping around.
Copying a shared garden
In the gardens panel, any garden shared with you has a copy button. It makes a real, independent copy in your own gardens, named "<original> (copy)". You're taken straight to your home canvas — the view pulls back and moves to the new node, which lands below your existing ones and flashes briefly so you can see where it went — every node inside becomes yours, with its own title, and editing it never touches the original. Copies keep their nested contents, connections and typed properties, and the types they use are matched to your own Types garden by name; a type you don't have yet is added for you. Two things are deliberately left behind: nodes shared with you but not opened to you copy across without their nested contents (you can't copy what you were never shown), and links pointing at nodes outside the copy become plain text rather than pointing back into someone else's garden. Anything the AI suggested but nobody accepted yet stays behind — those are the other person's pending ideas, not part of the garden. If a copied node has the same name as one of your own, it comes across numbered — "Health (1)" next to your "Health" — so you can tell them apart in search and the [[ picker. Images are the one thing a copy doesn't yet duplicate — they keep loading from the original, so they'll stop working if that garden stops being shared with you. The toast after a copy tells you exactly what came across.
Deleting and the trash
Deleting a node moves it to the garden's trash, where it can be restored. Guests can only delete nodes that have nothing nested inside them (or nodes shared with them directly), and the trash shows guests only their own deletions — restoring works the same way. Permanently deleting from the trash, including the Empty button, is reserved for the garden's owner.
Real-time presence
When multiple people work in the same garden, you see each other's cursors moving on the canvas in real time, each in a unique color. Presence is live on the shared garden's own canvas, and so are color regions and bookmarks there — a region someone draws, moves or renames shows up for everyone as it happens. Inside nested nodes, changes sync when saved rather than keystroke-by-keystroke, and regions on a nested canvas refresh on a short delay rather than instantly.
Roles and permissions
Owner: full control — manage members, restore or permanently delete anything, and delete the garden. Editor: create, edit, connect and delete on the canvases shared with them. Viewer: read-only. Removing a member removes all of their access inside the garden at once, including access to nodes they created.
Offline and sync
Neural Garden is local-first — you can work offline and changes sync automatically when you reconnect. If two people edit the same thought offline, changes merge gracefully when both come back online. An indicator in the UI shows your current sync status: a pending count while changes wait, "Sync paused" when the server can't be reached (or your session needs signing in again) — your changes stay queued and are retried — and "N unsynced" if the server refused a change outright. Click or tap the indicator for the details — the reason a change was refused, when the last sync happened — and for a "Sync now" button while changes are waiting and you are online. It is shown on phones too, and announces each change of state to screen readers.