Hosted projects: the staff list, and feedback across every project

/projects requires a real, signed-in aiKit account (the same accounts system /start/classic uses) — anyone without one is sent to log in first, and which projects a signed-in person actually sees is filtered further by their membership on each one. It's built for aiKit's own staff running client work, not for the client themselves — a client's own view of their project is the separate, password-protected Client Hub covered in its own guide, plus the plain feedback widget on their live site.

The dashboard is dark by default. The Light/Dark switch at the right end of its navigation bar (a sun or moon icon) is the same site-wide switch found in the header of every aiKit page — one choice covers the homepage, FAQ, Help, the /start/classic wizard, My Projects, and this dashboard alike. It's remembered in that browser and applied before the page paints, so it never flashes the other look on reload.

Every real hosted client site or app aiKit's team has linked in shows up here, one row (or tile) per project, with everything you'd otherwise have to open each project to check: whose court the ball is in, a real screenshot of the live homepage, where the site currently lives, and how much feedback is waiting.

The left rail lists every project you can see, always, grouped under its client (projects with no client sit under "No client yet" so nothing disappears), with sub-projects nested under their parent. It scrolls on its own when the list is long, and the box at the top of it ("Jump to a project…") narrows the rail as you type. Clicking a project name opens that project; the colored dot before it is the same whose-court signal as the row (waiting on us, waiting on client, all clear). Clicking a client name filters the main list to that client's projects; click it again to clear. The small chevron folds a client's projects away. A project with sub-projects carries a small count badge; pressing it filters the main list to that project and everything under it.

Below the rail, Kind lists every kind actually in use (Website, Landing page, Online store, App, Admin system, Dashboard, Client portal, Blog / content site, Marketing automation, Other) with counts. Stage says where each project is in its life: Being built, Built, not live, Live, Build failed, or Site is down (the automatic check could not reach it). How it started appears once you have projects from more than one origin: built from a brief, rebuilt from their site, from another AI tool, transferred in, cloned, or linked code.

Every filter lives in the page address, so a filtered view can be reloaded or shared. The "Showing N of M" bar names each active filter and "Show all →" clears them all.

"Advanced filters" in the toolbar opens a drawer above the list. Client and Project are dropdowns here as well as in the rail, so you can filter by a client AND one of its projects together (the Project list narrows to that client's projects once a client is picked; a project with sub-projects includes them). Next to those, three date ranges, each with a From and a To date: Created (when the project was added), Last change made (when the AI or a person last shipped a change), and Changes requested (when the client last left feedback or sent a request from the site's forms). Leave either end empty for an open range. Active dates show as chips in the "Showing" bar, the button carries a count of active ranges, and "Clear dates" removes them. Launch dates are not tracked yet, so there is no launch range.

Above the list, four chips filter by where a project stands right now, each with a live count and a colored dot: All, "Waiting on us" (a client's feedback or chat message needs a reply or real work), "Waiting on client" (a fix has shipped and the client hasn't confirmed it yet), and "All clear" (nothing open on either side). This status is worked out from real, current data — open feedback items, unanswered chat messages — not a flag anyone has to remember to set, so it can't quietly go stale.

The search box matches project name, company/client name, client email, and domain, live as you type. Sort has four options: "Needs attention first" (the default — waiting-on-us projects surface above waiting-on-client, then everything else by latest activity), "Latest activity," "Name A–Z," and "Newest created." A rows/grid toggle switches the whole list's layout — rows show more detail per project; grid is a denser, thumbnail-first view of the same projects.

The moment any filter is active — a search term, a status chip, a category, a client, a kind, a domain facet, or a date range — a bar appears above the list reading "Showing X of Y," with a small chip for each active filter (which status, which client, which category, and so on) and a "Show all →" button that clears every filter at once and goes back to the full list. Every one of these filters lives in the page's own URL, so a specific filtered view — "everything waiting on us for this one client," say — is a real, shareable, reloadable link, not just something you have to re-click each time.

The thumbnail on the left is a genuine screenshot of that project's actual live homepage (or, before it's live, a placeholder showing "Not live yet," "Capturing homepage…," or "Couldn't reach the site") — captured automatically once a day and refreshable on demand with the small refresh button in its corner. Clicking the thumbnail (or the project name) opens that project's own page.

Next to the name, a status pill repeats the same "Waiting on us / Waiting on client / All clear" wording as the filter chips. Below that: the client or company name (clicking it filters the whole list to that client), the kind of project, and its category/subcategory — each of these is itself a clickable filter, not just a label.

The domain line shows the live address (or, before launch, what's happening — "Build queued," "Importing their current site…," "Being built…," or "Ready to go live") next to a colored hosting badge in the same plain language as the Hosting-tab filter above. A small copy icon right next to a live address (2026-09-15) copies that exact address to your clipboard without opening the project — quicker than "Copy client link" in the row's menu when what you actually want is the real site URL itself. The badge itself is a link labeled "What does this mean?" that opens straight to that project's Hosting tab, and a "re-check" link next to it re-runs the domain check on demand without leaving the list.

An activity line names who did the last thing and when (client, staff, or the AI, each with its own icon) — the same signal the status pill is computed from.

Below that, the numbers that matter are real links, not static counts: "N need our work" and "N awaiting client review" both open the cross-project Feedback page pre-filtered to that exact project and status; "N shipped" opens that project's own Changelog tab; "N comments" opens the cross-project Feedback page showing every status for that project. Each row also carries "Open project →," "Chat" (opens the project with the Assistant's chat already open), "Live site ↗" (the live site with the feedback overlay turned on), and a "More ▾" menu with Requests, Version history, Brief & rules, Feedback, and "Copy client link" (a login-free link you can hand the client directly). The menu closes when you click anywhere else or press Escape. A project that has sub-projects lists them on its row as links ("2 sub-projects: …"), and a sub-project says which project it is part of.

Every row has a checkbox in the corner of its thumbnail. Tick two or more and a bar appears at the bottom of the list: "N sites selected · Ask the AI for N sites." That opens a dialog with one box: describe the change once, the way you would tell a designer ("add our holiday hours to the footer"), and press Run. The sites run one at a time, top to bottom. Each site gets exactly what a single chat edit gets — its own search of its own files, its own edits, its own new version you can undo, its own reload of the live site — so the result on each site is as accurate as doing it there by hand, and the cost is one AI edit per site (the dialog says how many before you run). While a site is being worked on you see the same working card as the Assistant's chat (clock, stage, the steps done so far). When a site finishes, its line shows what changed, the version number, a "view live" link, and a link to its versions.

A site that changed nothing, hit an error, was out of AI credits, or was already busy with another request is flagged "Needs a follow-up" with the reason and a link straight into that site's chat, so nothing is silently skipped. "Stop after this site" lets the current site finish and marks the rest Skipped. When the run ends, the list refreshes so the "Shipped:" activity lines and counts reflect the new versions.

Two buttons sit above the list itself: "New project" opens the Create-a-project wizard (its own guide), and "All feedback" jumps straight to the cross-project Feedback page below, already filtered to everything still pending.

At /projects/feedback sits every client note and staff note left on any hosted project, across all of them, in one filterable list instead of clicking into each project's own Feedback tab one at a time.

Status chips across the top — Pending (the default; everything not yet confirmed by the client), Open, In progress, "Fixed — awaiting client," Confirmed, and All — each carry a live count. A separate "Who filed it" toggle narrows to Everyone, Client, or Our team, and Client and Project dropdowns narrow further (the Project list itself narrows to just that client's projects once a client is picked). A search box matches message text, page, and submitter, and Sort offers Newest, Oldest, or Project A–Z.

Results are grouped by project, each group headed by the project's name (a link to that project's Feedback tab), its client/company name, and how many items are in the group. Each card underneath shows a status pill, a role tag (Client or Our team, with its own icon), a relative timestamp, the message itself, who submitted it, the page it's about, a reply count if there's an ongoing discussion thread, and whether a screenshot is attached. Clicking a card opens that exact item on its project's own Feedback tab — deep-linked and highlighted for a few seconds so it's unmistakable which one you followed the link to — and a separate "Open on live site ↗" link opens the actual live page the feedback is about.