The dashboard¶
Two interfaces: full dashboard and simple chat¶
Olano can show a person one of two entirely different screens after they sign in, chosen per account:
- Full dashboard — the complete control panel described in the rest of this section: every agent, the scheduler, files, memory, tools, connectors, configuration and diagnostics. This is the default for every account, including all existing ones, so nothing changes for anyone unless you change it deliberately.
- Simple chat — a stripped agent-chat app. A list of the agents that person has been given, the conversation with the selected one, attachments, voice, and the agent's profile. No settings, no diagnostics, no token counts, no model pickers. Suited to staff who only need to talk to an agent and would find the full console noisy.
People can also create agents of their own here: the New agent button in
the agent list (and on the you-have-no-agents screen) builds one from a template
or from a plain-words description. It asks for a name — a real name,
like "Hillary" — and shows the permanent handle it will generate from it;
a template's own label becomes the agent's title (shown under its name in
the profile and chat header), never its name. A Model dropdown lists only
the models this deployment can run right now — the deployment defaults, the
Olano tiers and any provider connected with its own key — pre-selected to
Default · Turbo; with a template picked, Template's own model keeps
the one it was designed for. Admins always have the button;
members have it while member bot creation is on, which it is by default —
see the self_serve_agents: block in config.yaml reference and
Config → Member bot creation.
By default the simple view shows replies only, with a single "working on it"
line while the agent acts. A detail button in the chat header cycles three
levels — Quiet, Normal (one compact line per tool the agent uses), Verbose
(each tool's full input and result, the same rendering the full dashboard
shows) — and typing /quiet, /normal or /verbose does the same. The
choice is remembered per browser. Typing /context shows the same
context-usage card the full dashboard's chat offers: how much of the agent's
working memory this conversation and its setup are using.
When the agent asks another agent for something, that "working on it" line is replaced by a more specific one — "maria is thinking…", with that agent's own face beside it — and it clears the moment the answer comes back. It shows at every level, Quiet included, because it is about who is working rather than about tools. Without it the chat looks frozen for as long as the other agent takes, which can be a while. What that agent says comes back inside this agent's reply rather than as a message of its own.
The agent list is built for real fleets. The selected agent's chats unfold under its row, each labelled with where it came from — SCHEDULED, TELEGRAM, WHATSAPP, AGENT and so on, the same badges the full dashboard's conversation list uses; an ordinary chat typed here carries no badge. A row of filter chips (All · Chats · Telegram · Scheduled · Agents · …) narrows the list to one source — the way to hide a chatty schedule's runs — and the choice is remembered per agent on that browser. The list under an agent shows only the most recent handful; "Show all N chats" opens the full list in the right-hand panel, with a search box, the same filter chips, and paging ("Show more") so an agent with hundreds of conversations stays fast. Selecting an agent opens its latest conversation — the newest one a person is in, typed here or arrived over a messenger, then the newest of any kind, and a fresh chat only when there is nothing at all (a fresh chat is not saved until you send something, so reopening an agent never piles up empty ones); switching agents and back keeps the chat you had open. A chevron at the end of every agent row folds or unfolds its chats in place: on the selected agent it is the fold, on any other it selects that agent and unfolds its chats without leaving the list. It shows on hover on a desktop (always on the selected row) and is always visible on a phone, where it is how you browse an agent's chats from the drawer — tap the chevron, then the chat you want; tapping the agent's name goes to its latest chat and closes the drawer. On a desktop, clicking the selected agent's row also folds its chats away (click again to unfold); right-clicking any agent opens a quick menu — new chat, see all chats, its profile, pin/unpin, move the row up or down, mark its unread as read; right-clicking a chat offers to rename, open or delete it. On a wide screen the whole list can be resized by dragging its right edge (double-click the edge to reset); the width is remembered per browser.
Arranging it. Each of the three headings — Pinned, All agents, Group chats — folds away on a click and stays folded on that browser, so a long list can be cut down to the part you use. Agents can be dragged up and down to the order you want, and dragged across the divider between Pinned and All agents to pin or unpin them; that order is remembered per browser too, and an untouched list stays alphabetical. Dragging is off while the search box has something in it (the order it would write covers agents the filter is hiding) — use Move up / Move down on the right-click menu instead, which is also the way to do it from a phone or a keyboard.
The button beside the Agents heading collapses the list to a strip of faces: names, previews, search and the open agent's chats go, the three headings stay, and clicking a face switches agents. Handy when a conversation wants the width; the same button brings the list back, at the width you had set.
Right-clicking the empty space below the list offers the two things this rail can create — New agent and New group — which is the fast way to either from anywhere in the list. (Right-clicking a row still offers that row's own actions; see above.)
A notification bell sits at the head of the list, the same one the full dashboard carries and fed by the same source: anything this deployment needs a person for — a channel that failed to start, a connector asking to be signed in, an action waiting on an approval — shows a count there and lists itself when you click it. View all opens the Overview panel below.
More. Below the group section, under a divider, a collapsible More group holds five of the full dashboard's pages — Overview, Insights, Issues, Cortex and Files — and opens them in place of the conversation. They are the same pages, not cut-down versions: Overview is the deployment snapshot with its notifications and approvals, Insights is usage and spend, Issues is the tracker, Cortex is what the agent has been thinking about, and Files is its workspace. Picking an agent, one of its chats, a group, or New chat while a page is open returns to the conversation (on a phone the drawer closes on it, as it does over a chat), and the browser's Back button brings the page back. A page a deployment has switched off in Config is not listed here either.
More starts closed on a browser that has never opened it, and after that it is whatever you left it — open it and it stays open across reloads, close it and it stays closed. Pages can also be dragged out of More into the space just above it, exactly as the full dashboard's own nav items can, so the one or two you actually use stay visible with More shut; drag them back in to undo it. Right-clicking a page offers the same move, which is how to do it from a phone or a keyboard. The arrangement is remembered per browser and is separate from the full dashboard's own — rearranging one does not disturb the other.
The account menu at the bottom of the agent list holds theme and language — one dropdown each — and, on a team install, Sign out; for an admin it also has the Simple view switch, which flips back to the full dashboard. It opens whole with the list collapsed to a strip too, hanging off the strip rather than being cut to its width.
Set it per person in Admin → Users, in the Interface column. On a
single-user install (no team accounts) the equivalent switch is the ui_mode
key in config.yaml — see config.yaml reference.
This is a display choice, not a permission. Moving someone to Simple chat does not restrict what they are allowed to do, and moving them to the full dashboard does not grant them anything. What a person may actually see and do is decided entirely by their role (admin or member) and by which agents they have been given access to, and those rules are enforced by the server whichever interface is on their screen. Use the Interface setting to reduce clutter; use Role and per-agent access to control access.
The full dashboard¶
The dashboard is the control panel for the whole deployment. The sidebar lists every section; you can reorder it by dragging, and the search box (top left) jumps to a section, an agent, a task, or an individual setting by name.
Finding a setting by name¶
The search box — or ⌘K / Ctrl+K from anywhere — also searches the deployment's settings, not just the list of sections. Type what the setting is called ("voice replies", "sandbox memory", "link lifetime", "deep model") and the results list the matching controls under a Settings heading, each showing the section it lives in (for example Config › Speech). Choosing one opens that page and scrolls straight to the control, highlighting it for a moment so you can see which of the page's many fields you were sent to.
Searching matches a setting's name, the section it sits in, common synonyms, and its help text — so a word you remember from the description finds it even when you don't recall the exact label. Several words narrow the results rather than widening them: "telegram allowed" finds only settings matching both.
Settings you are not allowed to change never appear: the whole Config page is admin-only, so a member account searching for "sandbox memory" gets no result for it rather than a link that then refuses them.
The sidebar has three parts. At the top, the search box and, right under it, the New agent button. Then your agents, the same list the simple chat shell shows: a Pinned heading, any groups you have made, All agents and, when group chats are on, Group chats. There is no separate agent dropdown: clicking an agent here selects it everywhere (the pages that are about one agent — Memory, Notes, Knowledge, Terminal — also carry their own picker in their top bar). Hover an agent to reveal its pin button (always visible on a touch screen): pinning keeps it under Pinned, on this device; click again to unpin. Right-click an agent for the rest — new chat, all chats, its page, move up or down, move to a group — and right-click anywhere else on the sidebar (a heading, the empty space, even the bar's foot) to make a new group to sort agents into. Below the agents, under a rule, a collapsible More group holds every day-to-day section: Overview, Insights, Cortex and Issues lead it, then the rest. Drag any tab to reorder More; that order is saved to your account, so it follows you to any browser you sign in from.
The setup and administration pages are not tabs at all. They live in the account menu — click your name at the bottom-left of the sidebar — in two groups: Setup holds Providers, Connections and Tools; Administration (admins only) holds Vault, Logs, Config and Admin. While one of those pages is open, your avatar carries the highlight the tabs would otherwise show. Every one of them is also a keystroke away in the ⌘K search.
The same menu holds Theme and Language, one dropdown each; both apply the moment you pick and are remembered per browser. Twelve themes. The two house themes lead the list, and the other ten follow dark-first:
| Theme | Page | Accent |
|---|---|---|
| Olano Dark (the default) | warm charcoal | Olano orange |
| Olano Light | neutral off-white | Olano orange |
| Midnight | deep navy | azure |
| Mocha | espresso brown | copper |
| Cool Black | pure black | green |
| Black & White | white | black — no colour but the states |
| Harbor | cool slate | deep teal |
| Sepia | warm parchment — nothing in it is pure white | umber |
| Sakura | blush pink | deep rose |
| Pastel Green | soft sage | emerald |
| Singapore | red blush, white cards | the flag's red |
| High Contrast | white | deep blue, with visible hairlines everywhere |
Cortex keeps its violet in every one of them. High Contrast is the accessibility choice rather than a taste one: body text is black on white and every status colour is pushed well past the usual readability floor. Sepia is the one to pick for long reading at night — no surface in it is pure white, so nothing on the page glares. The sun/moon button beside the bell at the top of the sidebar cycles through the same twelve.
There is no separate Chat tab: click an agent to open the chat with it (the full chat, with its mode, detail, model and reasoning controls). An agent's row shows a pulsing dot while it is working on a reply and a count of replies that landed while you were elsewhere. Type "chat" into the search box if you want the chat page by name.
Clicking an agent opens its latest conversation, not an empty one: the most
recent chat a person is in — typed in the dashboard or arrived over a messenger
such as Telegram or WhatsApp — wins, then the most recent thread of any kind (a
scheduled run, an exchange with another agent), and only an agent with no
history at all starts a fresh chat. A fresh chat is not saved until you send
something, so opening an agent again and again never litters its list with
empty conversations. Switching agents and back keeps the conversation you had
open. A copied /chat?thread=… link still opens exactly the thread it names.
The selected agent unfolds into its recent chats right under its name, the way the simple view's rail does: a New chat row, a row of filter chips (All, Chats for conversations typed in the dashboard, one chip per messaging app the agent talks on, Scheduled, Agents), the newest few conversations badged with where they came from, and Show all N chats — which opens the chat page's full list with search and multi-select delete. The chips only appear once the agent has more than one kind of chat, and the one you pick is remembered per agent on this device (the simple view reads the same choice). Click a chat to open it; right-click it for Open, Rename… (a name of your own, in place — Enter saves, Escape cancels; leave it empty to go back to the automatic title) and Delete chat (after a confirmation; admins of the agent only). The trash can on a hovered row deletes too. The chevron at the end of every agent row folds or unfolds its chats without leaving the page you are on — on a phone it is always visible, and it is how you browse an agent's chats from the drawer: tap the chevron, then the chat you want. Tapping the agent's name opens its latest chat instead.
Right-click an agent for its menu: New chat (a fresh conversation), Show / Hide its chats (the chevron's fold, for keyboards and long-presses), See all chats (the chat page with its conversation list open), About this agent (its page — profile, model, tools, channels), Pin / Unpin, Move up / Move down, Move to a group, and Mark N new as read when it has unread replies. Every item explains itself in its tooltip.
Groups are sections you make yourself to sort agents into — "Sales", "Personal". Right-click the empty space in the agent list (or any heading) and choose New group…, type a name and press Enter. Right-click a group's heading to Rename it or Delete it; only an empty group can be deleted — move its agents out first — and Pinned and All agents can never be. An agent is only ever in one place: pinning it takes it out of its group, and moving it into a group unpins it. Click any heading to fold that section away. Pins, groups and their order are remembered in this browser only; they change nothing about the agents themselves. With a row focused, the Up and Down arrow keys move through the list and Home and End jump to its ends.
Admins — and members, while Config → Member bot creation is on — also get a New agent button under the search box (or ⌘N / Ctrl+N). It is the same short flow as the simple shell's: a name, what the agent should be good at (a template, or from scratch), and which model it runs on. The model dropdown lists only what this deployment can run right now — the deployment defaults (Default · Turbo / Pro / Max), the Olano managed tiers (billed in credits, no key needed) and any provider you have connected with your own key — and comes pre-selected to Default · Turbo, the fast everyday tier. With a template picked, Template's own model keeps the model the template was designed for. The model can be changed later from the agent's page.
Four sections are admin-only and never appear for member accounts: Vault, Logs, Config, Admin. On a managed cloud deployment, other tabs can also be hidden by your plan's entitlements — if a section you expect is missing and you are an admin, that is why.
Installing it on a phone or tablet¶
The dashboard can be added to a phone's home screen and then behaves like an app: its own icon, full screen with no browser address bar, and the sign-in kept between launches. There is nothing to download — it is the same dashboard at the same address, and both interfaces (full dashboard and simple chat) work this way.
- iPhone and iPad. Open the dashboard in Safari, tap the Share button in the toolbar, scroll down to Add to Home Screen, and tap Add. Safari is the only browser on iOS that can do this. A prompt at the bottom of the screen offers the same steps the first time you visit on a phone; dismissing it is remembered.
- Android. Chrome shows an Install button in the same prompt, and also offers Install app in its own menu. Once installed, Olano appears in the Android share sheet: sharing a link or a note from another app opens the chat with that text already in the message box, waiting for you to add to it and send. Nothing is ever sent automatically.
- Desktop. Chrome and Edge can install it too, from the icon at the right of the address bar. There is no prompt for this — a browser tab is a perfectly good home for the dashboard on a computer.
An installed copy keeps working well enough to read the last screen you were on when the network drops, but it is not an offline app: agents run on the deployment, so sending a message needs a connection. When a new version of the dashboard has been deployed, a small A new version of Olano is ready bar appears with a Reload button; the update is applied when you choose it, so it can never interrupt a message you are part-way through writing.
Installing requires the dashboard to be served over https, which cloud
deployments always are. A self-hosted box reached over plain http (or by IP
address on a home network) cannot be installed — it still works normally in the
browser.
The installed app follows the phone's screen shape: on an iPhone it runs edge to edge, filling the screen from the top to the bottom edge, with the header clear of the status bar and the message box clear of the home indicator — no unused band at either end. It measures the screen it is actually on, so the same is true on every iPhone and Android size, on an iPad, and after rotating the device. Both the full dashboard and simple chat behave this way.
If a screen ever says "This view failed to load", press Reload on the card. That clears the app's stored copy of the old version and fetches the current one — it is the fix for a page that was left open across a deployment. If it keeps happening on the same screen, close the app entirely and open it again.