Skip to main content

Platform Tour (Screen by Screen)

This guide walks the entire dashboard, screen by screen, in roughly the order you meet it: the chrome that never leaves, the places you create things, the workspaces where the work happens, and the settings behind all of it. Each section says what you can do on that screen and links to the feature page that covers it properly.

Routes are written the way you would type them, without the locale prefix — the address bar shows /en/works, this guide says /works.

If you would rather read the concepts first, start with The Founder Journey; if you want the deep version of a single screen, follow the link at the end of each section.

1. Dashboard home — the cockpit at /

Signing in lands you on /. It opens with a welcome line, a row of stat tiles, and a stack of blocks. Three of those blocks are signal blocks: they render only when they have something to say, so a quiet account sees a shorter page rather than a row of empty shells.

BlockShownWhat you do there
Stat tilesAlwaysRead Missions, Ideas, Works, Items, Active Websites, Month Spend, Agents, Tasks in flight and Teams; click a tile to open its catalog.
Action approvalsWith pending proposalsApprove or reject a side-effectful action an Agent wants to take before it happens.
Needs attentionWith problemsJump straight to an errored Agent, a failed or paused schedule, a failed generation, a blocked Task, or an exhausted budget.
Coming upWith scheduled runsSee the next scheduled Work and Mission runs.
MissionsAlwaysPreview recent Missions with their Ideas / Works / Sites counters.
IdeasAlwaysSkim recent Ideas and press Refresh to have the platform research new proposals for you.
WorksAlwaysOpen a recent Work, or use + Add and View all.
TasksAlwaysOpen recent in-flight Tasks.
AgentsAlwaysCheck recent Agents, their status, scope and heartbeat.

Eight tiles are always there; Teams is a ninth that appears only once the Teams count can be read, so a workspace without it sees a row of eight rather than a tile stuck on zero. Month Spend is a link — it opens /settings/work-agent#account-budgets, where the account-wide cap is set. Agents and Tasks in flight each carry a qualifier underneath (n active, n blocked).

Every number on the page is fetched independently and every fetch is catch-defended, so a single failing call costs you one tile, never the page.

In depth: The Dashboard.

2. The chrome that never leaves

The sidebar

Thirteen entries, always in this order, with + New pinned above them and a collapse toggle at the top:

EntryRouteWhat lives there
Dashboard/The cockpit above.
Inbox/inboxQuestions, approvals, escalations and notices. Carries an unread badge.
Missions/missionsOngoing goals.
Goals/goalsMeasurable outcomes with a loop that advances them.
Ideas/ideasOne-shot proposals waiting to be accepted or dismissed.
Works/worksEverything you are building. Shows a pulse while a Work is generating.
Tasks/tasksThe task board and its triggers.
Teams/teamsPeople and Agents — the entry also covers /agents and /skills.
Memory/memoryOrg-wide knowledge; the entry also covers /meetings.
Templates/templatesWebsite, Work and Mission template catalogs.
Plugins/pluginsThe integration catalog.
Activity/activityThe audit log and the Schedules view.
Settings/settingsEverything in section 12.

Below the list sits the runner status pill — "Runner · n/N online". It appears only once at least one node is enrolled, and its popover shows each machine's status with a Refresh button. The bot button on the sidebar's outer edge opens the platform chat panel. Your avatar at the bottom opens the profile menu, in this order: Account Settings, Help & Docs, Support, Keyboard Shortcuts, Billing, Usage & Credits, Sign Out.

The organization switcher

At the top of the sidebar, the workspace switcher lists the organizations you belong to and offers Create Organization. Switching re-scopes the entire dashboard — Works, Agents, budgets, Memory — to that organization. See Organizations.

The header

ControlWhat it is for
Work switcherJump between your recent Works without going back to /works.
Setup badgeOnboarding {step}/{total} — appears when you closed the setup wizard early. Click it to resume exactly where you stopped.
Notification bellThe dropdown of what happened; each entry links to the thing it happened to. See Notifications.
Theme toggleLight / dark.
Help (?)Opens the Help drawer.

The Help drawer and keyboard shortcuts

The drawer has four tabs — Tips, Shortcuts, FAQ, Resources — plus an Onboarding card on the Tips tab whose button reads Open onboarding (n/N) and reopens the setup wizard where you left it. Resources link to the documentation, the GitHub repository, the issue tracker and community discussions, and the footer shows the running version and system status.

ShortcutDoes
Cmd/Ctrl + KSearch Works
CCreate a new Work
?Open the Help drawer

3. /new — the one place things get created

+ New in the sidebar opens /new: a prompt box and a row of chips that say what your prompt should become. Type a description (at least ten characters) and press Create.

ChipCreatesWhere it goes
MissionAn ongoing goalThrough the AI chat, then /missions/:id
IdeaA one-shot briefThrough the AI chat, then /ideas/:id
AgentAn AI teammate/agents/new (finish the dialog to create it)
TaskA trackable item/tasks/new
Website · Landing Page · Blog · Directory · Awesome RepoA Work of that kind/works/new?kind=…
CompanyA registered entity → an OrganizationThe company registration flow
StoreNothing yet — the chip is inertMarked Soon; see Store Builder

The default chip is Mission when you have none and Idea afterwards. /new?type=mission&template=<id> pre-fills the box from a Mission Template. Individual Work kinds can be switched off per installation, in which case their chip renders as Soon too.

In depth: The + New page.

4. The setup wizard

On a brand-new account the wizard opens by itself. It always has these ten steps, in order:

#StepWhat you decide
1WelcomeWhat Ever Works is; nothing to fill in.
2Your AI choiceEver Works AI, or bring your own key.
3Your Git StorageEver Works Git, your GitHub, GitLab, or your own Git.
4Your DB StorageThe managed database or your own.
5Your deploymentEver Works, Vercel, or Kubernetes.
6Where it runsHosted, your machine, or your own machines — this shapes the guidance the last step gives.
7What do you doYour roles and team size; used only to suggest starting points, and it hides nothing.
8CommunicationConnect Slack so you can mention the bot in a channel. Discord is marked Coming soon.
9Plugins & IntegrationsTurn on the plugins you already know you want.
10Create your first WorkGenerate now, and the platform builds it.

Three more steps appear only when your choices need them: Configure AI (any AI choice other than Ever Works), Connect your GitHub (Git storage set to your own GitHub), and Configure deployment (Vercel or Kubernetes). So the wizard is ten steps at its shortest and thirteen at its longest — which is why the badge shows a total rather than a fixed number.

Every step has Back, Skip step and Next, and progress is saved on the server as you go. Closing the wizard is safe: the header keeps a Setup badge, and the Help drawer's Open onboarding (n/N) button brings it back on the step you stopped at.

In depth: Onboarding & Setup Wizard.

5. Inbox

/inbox is the operator message center: everything addressed to you by your Agents, your Works and the platform — blocking questions, approval requests, escalations and notices — in one list with an unread badge on the sidebar.

How to answer a parked run:

  1. Open Inbox in the sidebar and stay on the Active view (the Archived view holds what you have already dealt with).
  2. Select the message. A question shows the agent's recommended option alongside the alternatives.
  3. Pick the recommended option, pick another option, or type a free-text answer.
  4. Press Send — this unparks the waiting run and seeds it with your answer.
  5. Archive the message when you are done. You can also mark messages read/unread, unarchive, or delete them.

In depth: Inbox · Approvals & Escalations.

6. Missions, Goals and Ideas

Missions

/missions lists your Missions with a prompt composer at the top (or create manually takes you to /missions/new), a status filter, search and pagination. /missions/:id gives you Run now, Clone, Pause / Resume, Complete, Delete and Save settings, plus the Mission's Ideas, its attached Works, its Goals panel and its spend against budget. The detail page carries its own tab strip — Overview | Tasks | Agents — at /missions/:id/tasks and /missions/:id/agents.

In depth: Missions.

Goals

/goals and /goals/new cover creation; /goals/:id is where a Goal is run. Lifecycle actions are Activate, Pause, Archive / Unarchive and Delete; Evaluate now scores progress on demand and lets you record an outcome. Below that sit the loop controls — Start, Advance, Nudge, Pause, Restart, Cancel and Adjust limits (including which Agent runs the loop) — with a progress sparkline, the event history and the sessions the loop spawned.

In depth: Goals.

Ideas

/ideas has the same prompt-first composer, a status filter (actionable or all, including done), search, and a gear menu holding the automation switches: Auto-generate, Auto-build, Auto-retry and Account budgets. On /ideas/:id you Accept an Idea to build it into a Work, Retry / Rebuild a failed build, Dismiss or Delete it, View Work once it exists, assign or unassign an Agent, and attach files. /ideas/:id/tasks and /ideas/:id/agents/new hang off the same Idea.

In depth: Ideas.

7. Works and /works/new

/works is the catalog: summary stats across your Works, the list itself, and search.

/works/new is the create form, and it has three modes:

ModeWhen to use it
AIDescribe the Work and let the agent prepare an approval-ready plan. The default.
ManualConfigure every step yourself — provider, template, items, deploy.
ImportBring in a Work you already have — repository, items and provider settings.

On the entry view you pick a kind chip — Website, Landing Page, Blog, Directory, Awesome Repo — and the form adapts: the picker below the chips merges your own website templates with the Blueprints published for that kind, and separate pickers choose the Git provider and the deploy provider. A Start a campaign chip hands off to /works/new/campaign for go-to-market Works.

Accepting an Idea sends you here with ?proposal=<id>, which pre-fills the form from the proposal and starts you in AI mode.

In depth: Creating a Work · Work Kinds · Work Blueprints · Work Import · Campaigns.

8. The Work workspace

Open any Work and you get a ten-tab workspace. The header above the tabs carries the repository links, the live-site link, the current status, and the Agents dropdown (pinned and assigned Agents, plus + New Agent).

TabRouteWhat you do there
Overview/works/:idStatus card, item/category/tag/comparison counts, budget summary, related Missions, and the Work's configuration at a glance.
Activity/works/:id/activityThe Work's feed with category and status filters, plus ingested connector events — commits, PRs, tracker items, docs, chat, meetings.
Items/works/:id/itemsFive sub-tabs: Browse items, Categories, Tags, Collections, Source health. Add, edit and bulk-edit items, import and export them, manage taxonomy, and set screenshot and validation options.
Tasks/works/:id/tasksThe task list filtered to this Work, with its run summary.
Pull requests/works/:id/pull-requestsOpen PRs across the Work's repositories: refresh, open, view the diff, Request review by the agent, and read the review history.
Memory/works/:id/kbThe Knowledge Base workbench — a document tree, upload zone, classification, a markdown editor with wikilinks and mentions, history, locks and citations, plus a Cmd/Ctrl+K search palette and a side AI panel. There is no create-document button: uploading and the chat rail are the two browser paths (see Knowledge Base and Memory).
Worker/works/:id/generatorWhere generation happens — see the sub-tabs below.
Plugins/works/:id/pluginsEnable or disable plugins for this Work and pick which one provides each capability.
Deploy/works/:id/deployChoose the deploy provider, deploy or update the repository, watch progress, and manage domains, subdomains and runtime environment variables.
Settings/works/:id/settingsThree sub-tabs: General, Members, Budgets & Usage.

Tabs are permission- and kind-aware, so the strip is shorter for some Works and some roles. The Items tab is labelled Posts on a blog and Pages on a website, and disappears entirely for a landing page; Worker, Deploy, Plugins and Settings appear only if you may generate, deploy or administer this Work.

Worker sub-tabs

Sub-tabRouteWhat you do
Generate/works/:id/generatorStart generation, Recreate Work, or Update items — choose a provider, fill plugin fields, open advanced options, and Cancel a run in flight while progress streams.
Schedule/works/:id/generator/schedulePut generation on a cadence and override the providers a scheduled run uses.
History/works/:id/generator/historyPast runs and what each one changed.
Comparisons/works/:id/generator/comparisonsGenerate and edit comparison pages; each opens at /works/:id/generator/comparisons/:slug.

Settings sub-tabs

Sub-tabRouteWhat lives there
General/works/:id/settingsWebsite configuration, repository visibility, provider repositories, committer identity, community-PR handling, merge policy, quality gates, task isolation, advanced prompts, sources, external references, activity sync, item import/export, the README configuration — and deletion.
Members/works/:id/settings/membersInvite people, set roles, transfer ownership. Visible only if you may manage members.
Budgets & Usage/works/:id/settings/budgets-usagePer-plugin caps and the spend summary for this Work.

In depth: Items · Knowledge Base · Scheduled Updates · Comparisons · Merge Policy · Quality Gates · Custom Domains · Works Config · Work Members · Budgets & Usage.

9. Tasks

The hub

/tasks carries a two-entry tab strip — Tasks | Triggers — and three views of the same list: Cards, Table and Kanban. Filter by status (backlog, todo, in progress, in review, blocked, done, cancelled), by priority p0p4, by label, or by free-text search; the status filter hides itself in Kanban, where the columns already group by status. Run with agent starts a run from the list, including in batch.

  • /tasks/new — the create form.
  • /tasks/templates — reusable task and workflow templates.
  • /tasks/triggers and /tasks/triggers/:id — inbound webhooks that turn an external event into a Task: copy the webhook URL and a signed curl example, rotate the secret, Fire now to test, Pause, and read the recent fires.

Task detail

/tasks/:id is a two-column issue view. The main column holds the title, the status buttons (only legal transitions are offered), the description editor, attachments and the activity thread; the right rail is a sticky Details panel with status, priority, labels, dates and scope.

What you can do from here:

  1. Transition the task through the workflow, or Delete it.
  2. Edit the description inline with EditSave.
  3. Talk to it — post a chat message, mention an Agent (@ceo can you review this?), or start a message with / to invoke a Skill.
  4. Re-scope it — attach or detach the Work, Mission, Idea and Agent it belongs to.
  5. Run itRun with agent, then Steer, Interrupt or Resume the run, and read the run history.
  6. Review the code — the branch section, the diff sheet, the PR pill and the checks section show what the run produced; you can discard the branch or resolve decision conflicts.
  7. Repeat it — subtasks, a recurring schedule, isolation on or off, and a maximum attempt count.

In depth: Tasks · Inbound Triggers · Task Isolation · Sessions & Run Steering.

10. Teams, Agents and Sessions

One hub with four tabs — Teams | Agents | Sessions | Archived — because people and Agents are one organization seen through two doors.

Teams

/teams lists your teams. /teams/new asks for a parent team and a manager Agent. /teams/:id is the roster: add an Agent or a person as Lead or Member, remove them, nest sub-teams, and see the Works, Tasks, Agents, Missions and Ideas the team owns. /teams/:id/settings re-parents the team, changes the manager or deletes it. /teams/org-chart draws the whole hierarchy, humans and Agents together.

Agents

The Agents tab is prompt-first: describe the teammate you want, or pick one of the template chips (View All opens the full catalog, and your own templates appear alongside the shipped ones). Create Agent Manually opens /agents/new, where you choose a template, a scope (Tenant, Mission, Work or Idea), a Team and a reports-to Agent. The header's Agents Chart button opens /agents/chart — the org chart with human members stripped out. Below the list sits the Skills block; /skills redirects there, and /skills/new, /skills/:id and /skills/templates are its own pages.

The Agent workbench

/agents/:id has ten tabs:

TabWhat you do
DashboardStatus, heartbeat, idle behaviour, last and next run, capabilities, attachments and the scorecard.
ActivityEverything this Agent has done.
TerminalThe live run stream; ?run=<id> opens a specific run.
InstructionsThe five definition files — SOUL.md, AGENTS.md, HEARTBEAT.md, TOOLS.md, agent.yml — with autosave.
SkillsBind and unbind Skills.
CapabilitiesTool grants, skills, MCP access, repositories, environment, init script and permission flags.
MCP ServersThe MCP servers this Agent may call.
CollaboratorsThe sub-agents it is allowed to delegate to.
BudgetsIts spend cap and what it has spent.
SettingsHeartbeat cadence, idle behaviour, pause-after-failures, scope, and archiving.

An Agent with a mailbox also has /agents/:id/inbox, with /agents/:id/inbox/compose to write and /agents/:id/inbox/:messageId to read a thread.

Sessions and Archived

/agents/sessions lists runs across every Agent, filtered by status or Work, grouped by Work, with a badge on the ones awaiting input. /agents/sessions/:runId shows the tool calls with their arguments and results, and lets you Steer, Interrupt or Cancel the run — and jump to the Agent's Terminal tab for that run via Open terminal when the run has a terminal. /agents/archived is where archived Agents go; restore them, or delete them permanently.

In depth: Teams · Agents · Agents Catalog · Agent Capabilities · Live Agent Terminals · Agent Email · Sessions & Run Steering · Skills Catalog.

11. Memory, Templates, Plugins and Activity

Memory — /memory

Every Work's Knowledge Base plus your organization-level documents in one searchable list, faceted by type, Work, source and status. The page also holds the review panel (accept or reject agent-written material), the uploads and files panels, the consolidation preview you can inspect before applying, its cadence and mode settings, and the Meetings panel at /memory#meetings. /meetings/new takes a pasted transcript and routes it to a Work; /meetings/:id shows the summary and transcript and lets you re-route it.

In depth: Memory · Decisions & the Memory Review Queue · Meetings.

Templates — /templates

One catalog, three kinds: ?kind=website, ?kind=work and ?kind=mission. Filter by All / Built-in / Custom, Add template from a repository URL (framework, branches, preview), Customize template with an AI provider, Fork a Work template into your GitHub user or organization, and Sync a custom template with its base or turn on auto-update. Mission templates offer Use this, which drops you back on /new with the prompt pre-filled.

In depth: Website Templates · Work Templates · Mission Templates.

Plugins — /plugins

Search the catalog, filter by category or capability, and enable or disable a plugin for the whole account; the install state is shown on each card. /plugins/:pluginId is the plugin's own page: enable it, fill in its settings form, connect it via OAuth or device authorization, and read its README and documentation link.

In depth: Plugins.

Activity — /activity

A segmented view of Log and Schedules. The Log renders as cards, a table or a board, filtered by action type, status and search, with pagination and Export CSV. Your view choice is remembered locally, and ?view=schedules is a shareable link straight to the schedules side.

In depth: Activity.

12. Settings

/settings is a page with its own sidebar: a fixed list of account and platform entries, then a PLUGINS section built from the plugins your installation actually has, then Danger Zone pinned at the bottom.

EntryRouteOne line
Profile/settingsName, username, avatar, budget-alert email, and email verification.
Organization/settings/organizationThe organizations you belong to and the Vision set on one.
Security/settings/securityChange your password.
API Keys/settings/api-keysKeys for the REST API and the CLI.
Data/settings/dataExport your account as JSON, import it back, sync configuration to a private repository.
GitHub App/settings/github-appInstallations linked to this workspace; re-sync the snapshot and onboard a repository.
Repositories/settings/repositoriesThe account-level repository registry Agents can be granted, including Work-derived repos.
Work Agent/settings/work-agentThe built-in agent's preferences, its build requests, and its active run.
Fleet/settings/fleetEnroll a node, mint tokens, scan the QR code, set execution preferences.
Job Runtime/settings/job-runtimeWhether jobs use the platform runtime, your credentials, or a provider you pick.
Environments/settings/environmentsThe sandbox an Agent's work runs in — packages and networking.
Connections/settings/connectionsThe registry of external MCP servers your Agents may consume.
Digest/settings/digestPersonal and organization briefing cadence; the two are separate settings.
Notifications/settings/notificationsThe event-by-channel preference grid and the in-app inbox.
Channels/settings/integrations/channelsCreate and manage the channels notifications are delivered through.
Email Addresses/settings/integrations/emailsTenant-managed inbound and outbound addresses for Agents.
Billing/settings/billingPlan, balance, invoices, credits ledger, and /settings/billing/payment-method.
Usage & Credits/settings/usageWhere the credits went — per day, model, agent and Work, with Export CSV. ?tab=costs opens the costs dashboard.
Plugins by category/settings/plugins/:categoryConfigure the plugins of one category; a dot marks a category with required settings missing.
Danger Zone/settings/dangerExport everything, then delete the account behind a typed-email confirmation.

Fleet sits directly above Job Runtime on purpose: Fleet is where work can run, Job Runtime is how it is dispatched. The Fleet entry disappears entirely on installations where the operator turned Fleet off.

Three admin pages exist for platform operators only — /admin/usage, /admin/plugins/allowlist and /admin/tenants/:tenantId/runtime-allowlist — and they return a 404 for everyone else, so their absence from your sidebar is by design rather than a missing permission.

In depth: The Settings Map · API Keys · Data Management · Repositories Registry · Fleet · Job Runtimes · Environments · MCP Connections · Digests · Notifications · Credits & Billing.

13. The same screens without a browser

Every screen in this tour is a view over the same REST API, so anything you can click you can also script — with your own permissions, not a weaker service account. Three surfaces reach it:

SurfaceWhat it isReach for it when
CLI (ever-works)The public command-line client, with auth, work, kb and plugins.You live in a terminal, or you are wiring the platform into CI.
MCP server74 tools — 66 generated from the API's OpenAPI spec plus the kb.* Knowledge Base tools, register_work and ping — over stdio or streamable HTTP.You want an AI assistant to drive the platform in conversation.
REST APIThe endpoints every screen above already calls.You are writing your own integration or automation.

All three authenticate with an API key you mint on Settings → API Keys (/settings/api-keys).

How to walk the Works half of this tour from a terminal:

  1. ever-works auth login --api-url https://api.ever.works — start a session. On a headless box add --manual and paste the key from /settings/api-keys instead of opening a browser.
  2. ever-works work list — the terminal equivalent of /works.
  3. ever-works work create — an interactive walk through what /works/new asks in Manual mode.
  4. ever-works work generate — the Start generation button on /works/:id/generator.
  5. ever-works work status — the progress panel on that same tab.
  6. ever-works work deploy — the Deploy tab at /works/:id/deploy.
  7. ever-works kb list <workId> and ever-works kb upload <workId> ./brand-guide.pdf --class brand — the Memory tab at /works/:id/kb.

The endpoints behind the busiest screens:

ScreenEndpoint
/worksGET /api/works (supports ?search= and pagination)
/works/newPOST /api/works
Worker → GeneratePOST /api/works/:id/generate
Worker → HistoryGET /api/works/:id/history
Work → MemoryGET /api/works/:id/kb/documents
Teams → AgentsGET /api/agents, POST /api/agents
An Agent's Pause / ResumePOST /api/agents/:id/pause, POST /api/agents/:id/resume
/tasksGET /api/tasks (status, priority, scope, label, search)
A Task's status buttonsPOST /api/tasks/:id/transition
A Task's chat threadGET and POST /api/tasks/:id/chat

In depth: CLI Commands · KB MCP & CLI Reference · MCP Server · API Authentication.

14. What is not finished yet

A tour is only useful if it is honest about the edges. As of today:

ScreenState
Store chip on /new and /works/newInert, labelled Soon. The kind is flag-controlled like every other kind, and the flag is off. See Store Builder.
Discord on the Communication wizard stepMarked Coming soon; Slack is the shipped connector.
Settings → Email AddressesA first-pass shell: the list renders, but Add address opens a placeholder rather than the full add wizard.
Settings → Notifications preference gridThe event-by-channel matrix renders read-only in this version; in-app delivery is always on, and channel routing is edited from Channels.
Admin pagesPlatform-operator only, and a 404 for everyone else.
Billing purchasesThe purchase, payment-method and auto-recharge cards sit behind a server-side switch that is off by default, and need a configured payment provider; otherwise the page shows a coming-soon state instead of buttons that would error.