Aller au contenu principal

The Dashboard

Sign in and you land on the Dashboard — the home page at /. It is a cockpit, not a launcher: one screen that tells you what needs a decision right now, what is about to run, and how the Missions, Ideas, Works, Tasks and Agents you own are doing. Around it sit the two pieces of chrome that stay with you on every page — the sidebar on the left and the header on top — plus the platform chat panel, which opens from the sidebar's right edge.

This page is a tour of all three: what each block shows, when it appears, and which route every button and tile leads to.

The home page at /

The page opens with "Welcome back, {username}!" and a strip of stat tiles, then a stack of blocks. Three of the blocks are signal blocks — they render only when they have something to say, so a healthy, quiet account sees a shorter page, not empty shells.

BlockAlways shown?What it is for
Stat tilesYes (Teams tile is conditional)Counts and month spend, each a shortcut to its catalog
Action approvalsOnly with pending proposalsApprove or reject side-effectful actions your Agents want to take
Needs attentionOnly with problemsErrored Agents, failed builds, blocked Tasks, an exhausted budget
Coming upOnly with scheduled runsThe next three scheduled Work or Mission runs
MissionsYesThree most recent Missions with Ideas / Works / Sites counters
IdeasYesThree Idea cards, quick-add, AI research ("Suggest more")
WorksYesSix most recent Works
TasksYesFive most recent in-flight Tasks, list or card view
AgentsYesThree most recent Agents with status, scope and heartbeat

Every number on the page is fetched in parallel when the page loads, and every fetch is caught individually: if one backend call fails you see a zero or a missing block, never a broken home page.

The nine stat tiles

Each tile is two lines — the icon and count, then the label with an optional qualifier in parentheses. Tiles with a link open the matching catalog.

TileWhat it countsClick goes to
Total MissionsMissions you own/missions
Total IdeasIdeas of every status/ideas
Total WorksWorks you own or can access/works
Total ItemsItems across all your Works
Active WebsitesWorks that are deployed and live
Month SpendAccount-wide spend for the current month, in your billing currency/settings/work-agent#account-budgets (the account cap)
AgentsAll Agents, with (n active) inline/agents
Tasks in flightTasks in todo, in progress or in review; (n blocked) appears only when something is blocked/tasks
TeamsTeams across your Organizations/teams

The Teams tile only appears once you belong to at least one Organization; with none, the grid stays at eight tiles. The Works, Items and Active Websites numbers come from GET /api/works/stats, the same endpoint the /works page uses for its summary. Month Spend is the number Budgets & Usage caps; the tile is a shortcut to that cap.

Needs attention

The red-and-amber Needs attention block collects the things that will not fix themselves. Each card is a link straight to the entity, ordered danger first, then most recent first, and capped at six on the home page — the rest stay discoverable on their own catalog pages.

Card titleFires whenSeverityOpens
Agent "{name}" erroredAn Agent's status is error — it auto-paused after repeated failures; the badge shows its error countDanger/agents/:id
Generation failedAn Idea's build failed ("{name}" could not be built)Warning/ideas/:id
Task blockedA Task is in blocked and needs input to continueWarning/tasks/:id
Budget exceededAccount-wide spend has reached its capDanger when the cap hard-stops runs, warning when spend is over the cap but still allowed/settings/work-agent#account-budgets

Two more card kinds — Schedule run failed and Schedule paused — are designed into the block, but the home page does not yet compose them; today a failed or paused schedule is visible in the Schedules view of Activity (/activity?view=schedules) and on the Work's own Schedule tab (see Scheduled Updates).

Coming up

Coming up lists the next three scheduled runs across your account — a Work schedule or a scheduled Mission tick — soonest first, each row showing the Work or Mission name, a WORK / MISSION chip, and the next run date and time. When more runs are queued, a +N more link opens the full Schedules view at /activity?view=schedules.

The rows come from GET /api/schedules?enabledOnly=true, the same aggregation the Schedules view reads, so the two never disagree. A schedule with no computable next run is not "coming up" and is left out; an account with nothing scheduled sees no block at all.

Action approvals

When an Agent wants to do something side-effectful that its permissions do not let it do on its own, it files a proposal and waits. Those proposals surface at the top of the Dashboard as Action approvals, with a count badge and the note "Your agents proposed these side-effectful actions. Approve or reject each one."

Each row shows:

  • the action typeSpawn agent, Schedule task, Send message, Budget override, or a generic Action;
  • any risk flagsBudget override, Destructive, Cross-scope, High fan-out;
  • the proposal's title;
  • Reject and Approve buttons. Approve all in the block header approves every pending row in one call; rows decided elsewhere in the meantime are skipped and reported ("Approved 3 actions, 1 already decided").

The block disappears as soon as the last row is decided. The same queue is available over the API — GET /api/agent-approvals (defaults to the pending queue; filter with ?status=), POST /api/agent-approvals/:id/approve, POST /api/agent-approvals/:id/reject, and POST /api/agent-approvals/approve-all with an optional ids list. Deciding an already-decided proposal returns 409.

Approval requests also land in the operator Inbox at /inbox (the second sidebar entry), next to the blocking questions and escalations your Agents send you, so you can answer them from either place.

Missions

The Missions block previews your three most recent Missions. Each card links to /missions/:id and shows the title, a status pill, a Scheduled badge for Missions that run on a cadence, and three counters: Ideas (every Idea attached to the Mission, any status), Works (Ideas that were accepted and became Works) and Sites. The Sites counter is reserved for the per-Mission deployment count and currently reads 0 for every card.

  • + Add opens the unified picker at /new?type=mission.
  • View all (N) → opens the /missions catalog.
  • With no Missions yet, the block shows "No Missions yet. Start one to give the agent a persistent Goal that keeps generating Ideas for you."

See Missions.

Ideas

The Ideas block is the home-page window onto the Idea queue described in Ideas. It shows up to three Idea cards — each with Build, View Work (once built) and Dismiss — and a header row of controls:

ControlWhat it does
Show accepted / Show dismissedToggles that filter the preview exactly like the /ideas page. On the home page both start on, so an Idea you created by hand is visible whatever its status.
+ AddOpens an inline text box. Describe the Idea (at least 10 characters) and press Add — the prompt is handed to the platform chat, which creates the Idea for you. Create manually beside it opens the deterministic form at /ideas/new.
Settings (gear)Shortcuts into Settings → Work Agent: Auto-generate Ideas…, Auto-build Works…, Auto-retry policy…, Account-wide budgets….
Suggest more (refresh icon)Asks the platform to research new Ideas for you. The block shows "Generating ideas…" and polls every 2.5 seconds (for up to ten minutes) until the research run finishes. Hitting the daily limit shows "Daily limit reached. Try again tomorrow."
View all (N) →Opens /ideas; N is the real total across every status, the same number as the Total Ideas tile.

Over the API, Suggest more is POST /api/me/work-proposals/refresh; GET /api/me/work-proposals/status reports whether research is running and whether another refresh is allowed, and GET /api/me/work-proposals lists the Ideas.

Arriving at the Dashboard with ?newUser=true and an empty Idea queue starts a research run automatically — this is what a brand-new account sees after the onboarding wizard hands it over.

Works

Works lists your six most recent Works as cards, with + Add opening the picker at /new?type=website. Once you have more than five Works, View all (N) opens the /works catalog. An account with no Works sees "No works yet" with a Create Your First Work button that opens the same picker. See Creating a Work.

Tasks

Tasks shows your five most recent in-flight Tasks — anything in todo, in progress, in review or blocked. Each row links to /tasks/:id and carries the Task slug, its title, a priority pill (p0p4) and a status pill.

  • The list / cards toggle switches between a compact list and a card grid; the choice is remembered in your browser.
  • + Add opens /tasks/new.
  • View all (N) → opens /tasks; N is the Tasks-in-flight count from the tile.

See Tasks.

Agents

Agents previews your three most recent Agents. Each card links to /agents/:id and shows the avatar or initials, name, title, a status chip (draft, active, paused, running, error, archived), the scope (Workspace, Mission, Work or Idea) and, for Agents with a heartbeat, every {cadence}.

  • + Add opens /agents/new.
  • View all (N) → opens the Agents tab of the Teams hub at /agents.
  • With no Agents yet: "No Agents yet. Create your first Agent to delegate recurring work."

See Agents.

The sidebar

The sidebar is the same on every dashboard page. Top to bottom:

Organization switcher

The very first control is the Organization switcher — the active Organization's avatar and name (or the Ever Works wordmark when you have no Organization yet) with a chevron. Click it to open Switch Organization: every Organization you belong to, a check mark on the active one, and Create Organization, which opens the creation dialog. Picking another Organization persists the choice on your account and reloads the Dashboard scoped to it. Organizations, invitations and roles are covered in Teams & Organizations.

+ New

The primary + New button opens the unified picker at /new, where you choose what you are starting — a Mission, an Idea, an Agent, a Task, or a Work of a given kind (Website, Landing Page, Blog, Directory, Awesome Repo, Company) — and describe it in one prompt. Every chip, and where each one leads, is described on the + New page.

EntryRouteNotes
Dashboard/This page.
Inbox/inboxMessages addressed to you: blocking Agent questions, approval requests, escalations, notices. The only entry with an unread badge (refreshed every 30 seconds).
Missions/missionsSee Missions.
Goals/goalsSee Goals.
Ideas/ideasSee Ideas.
Works/worksShows a small amber activity dot while a Work is generating and you have not visited /works yet.
Tasks/tasksSee Tasks.
Teams/teamsThe hub for people and Agents — tabs Teams | Agents | Sessions | Archived. Stays highlighted on /teams/*, /agents/* and /skills/*.
Memory/memoryOrganization-wide memory; also hosts the Meetings block, so /meetings/* keeps it highlighted. See Decisions & Review.
Templates/templatesWebsite, Work and Mission template catalogs. See Work Templates.
Plugins/pluginsSee Plugins.
Activity/activityThe log and the Schedules view. See Activity.
Settings/settingsSee The Settings Map.

Collapse toggle

The small panel icon beside the switcher collapses the sidebar to an icon-only rail (hover an icon for its name) and expands it again. The state is remembered in a cookie, so it survives reloads. On narrow screens the sidebar becomes an overlay: open it from the menu button at the left of the header, close it with the X or by tapping outside.

Runner status

Above the profile menu sits the Runner pill — but only once your account has at least one enrolled Fleet node. Until then the footer is exactly as it was, so an account that never runs work locally never sees it.

The pill reads Runner · Running / Busy / Offline / None with N of M online underneath. Click it for the Local runners popover: one row per machine (laptop icon for a desktop node, server icon for a fleet node) with its state — Online, Busy, Offline, Paused, Disabled or Enrolling — and four facts that explain a runner that looks healthy but is not: Last heartbeat, Daemon version, Agent CLI version (or Not installed) and Disk free. Refresh now re-polls; the caption tells you the automatic cadence (Refreshes every Ns); Manage runners opens /settings/fleet.

Profile menu

Your avatar, username and email at the bottom open the profile menu:

ItemGoes to
Account Settings/settings
Help & Docsdocs.ever.works (new tab)
SupportThe GitHub issue tracker (new tab)
Keyboard ShortcutsOpens the Help drawer
Billing/settings/billing — see Credits & Billing
Usage & Credits/settings/usage — see The Settings Map
Sign OutEnds the session

The chat panel

When the platform chat is collapsed, a small expand button sits on the sidebar's right edge. Open it and the chat becomes a resizable side panel (drag the grip to resize, use the arrows to collapse it again or expand it to fill the screen); on phones it opens full-screen. Its width and open state are remembered between visits. What you can ask it to do — which is nearly everything the dashboard does — is covered in Platform Chat.

The header

Left to right, the header carries:

ControlWhat it does
MenuOn narrow screens only — opens the sidebar overlay.
Work switcherAppears only while you are inside a Work (/works/:id/…). A search box that filters your Works by name or slug; pick another one and you land on the same tab of that Work, query string included. The badge beside the name shows the Work's generation status — it shimmers while generating and refreshes every five seconds; a warning icon means the last run finished with warnings, and hovering it shows the message.
Setup badgeOnboarding n/N — appears when you closed the setup wizard before finishing and have no Works yet. Click it to reopen the wizard where you left off; the X ("Hide onboarding shortcut") hides it in this browser without marking setup complete. N is 7 to 10 depending on which steps your choices require.
NotificationsThe bell, with an unread count (99+ when it overflows), refreshed every 30 seconds. The dropdown lists each notification with its type icon, message, relative time and, when it has one, an action link; clicking a notification marks it read and follows the link. X dismisses a notification that is not persistent, and mark all as read clears the count. Empty state: "No new notifications".
Theme toggleSwitches between light and dark mode.
Help (?)Opens the Help drawer described below.

Directly beneath the header, a self-hosted install without a background job runtime shows the banner "Background job runtime is not configured. Agent runs cannot execute on this install until a job runtime (e.g. Trigger.dev credentials) is set up." with a Setup guide link and a Dismiss button. Configuring one is covered in Workers.

The Help drawer

Help & Resources slides in from the right — from the ? button in the header, the Keyboard Shortcuts entry in the profile menu, or by pressing ? anywhere outside a text field. It has four tabs:

TabContents
TipsAn Onboarding card — Open onboarding (n/N) reopens the guided setup and continues where you left off — followed by four Quick Tips (create your first Work, use the AI generator, connect GitHub, press ? to reopen this panel).
ShortcutsThe keyboard shortcut table below, with the hint "Use Cmd on Mac or Ctrl on Windows/Linux".
FAQThree expandable answers — How does AI generation work?, How do I connect GitHub?, Can I edit items after generating? — each with numbered steps and a Learn more link into the documentation.
ResourcesLinks to Documentation, the GitHub Repository, Report an Issue and Community Discussions; a Need more help? card with Join the discussion; and a System card showing All systems operational (with a status-page link when the operator has configured one) and the running Environment.

The drawer's footer shows the running version (Ever Works v…).

Keyboard shortcuts

Exactly three shortcuts are bound, and they work on every dashboard page:

KeysActionWorks inside a text field?
Ctrl + K / ⌘KGo to /works?focus=search — the Works catalog with the search box focusedYes
CStart a new Work — opens /works/new, which forwards to the unified /new picker unless a mode or an Idea (?proposal=) is givenNo
?Open the Help drawerNo

Ctrl/ + K deliberately fires even while you are typing, so you can jump to search from anywhere; C and ? are ignored inside inputs, text areas, selects and editable content so they never hijack what you are writing.

How to

Triage your account in two minutes

  1. Open /. If Action approvals is showing, read each proposal's action type and risk flags, then click Approve or Reject — or Approve all when the whole queue is routine.
  2. Work through Needs attention top to bottom; every card opens the Agent, Idea, Task or budget page where the fix happens. A danger card (an errored Agent, a hard-stopped budget) is worth opening first.
  3. Check the Inbox badge in the sidebar; an unread count means an Agent is parked waiting for your answer. Open /inbox, pick the recommended option or write a reply, and Send Reply.
  4. Glance at Coming up to see what will run next and when; +N more shows the whole schedule.
  5. Read the Month Spend tile against your cap; click it to adjust the account-wide budget.

Get a fresh batch of Ideas

  1. In the Ideas block, click the refresh icon (Suggest more).
  2. Wait for "Generating ideas…" to finish — the block polls on its own and fills in as soon as the research run completes.
  3. On each new card, click Build to turn it into a Work, or Dismiss. Show dismissed brings dismissed Ideas back if you change your mind.
  4. To make this recurring, open the gear menu and choose Auto-generate Ideas…, which takes you to the setting in Settings → Work Agent.

From a terminal, the same run is one call with your API key:

curl -X POST http://localhost:3100/api/me/work-proposals/refresh \
-H "x-api-key: YOUR_API_KEY"

curl http://localhost:3100/api/me/work-proposals/status \
-H "x-api-key: YOUR_API_KEY"

Approve a proposal from the API

  1. List the pending queue: GET /api/agent-approvals (add ?status=approved or ?status=rejected to review past decisions).
  2. Decide one row with POST /api/agent-approvals/:id/approve or POST /api/agent-approvals/:id/reject. A 409 means someone already decided it.
  3. Or clear the queue in one call: POST /api/agent-approvals/approve-all, optionally with { "ids": [...] } to approve a subset.
curl http://localhost:3100/api/agent-approvals \
-H "x-api-key: YOUR_API_KEY"

curl -X POST http://localhost:3100/api/agent-approvals/<proposal-id>/approve \
-H "x-api-key: YOUR_API_KEY"

Switch to another Organization

  1. Click the Organization switcher at the top of the sidebar.
  2. Pick an Organization from Switch Organization — the check mark shows which one is active now.
  3. The Dashboard reloads scoped to that Organization; the Teams tile, the Missions and Works you see, and the Memory page all follow it.
  4. Need a new one? Choose Create Organization from the same menu.

Pick up onboarding where you left off

  1. If the Onboarding n/N badge is in the header, click it.
  2. Otherwise press ? (or click the header's ? button), stay on the Tips tab and click Open onboarding (n/N).
  3. The wizard reopens on the step you stopped at; finish it or close it again — closing never discards the choices you already saved.