Types of chats
Selecting an agent via
@ in the composer or the ”+” menu when creating a chat turns it into an agent chat.
The Home screen
“New chat” in the sidebar opens the workspace’s Home: three entry cards, a “Quick access” row beneath them, and the chat box docked at the bottom of the window (“What would you like to accomplish today?”). Type in the box to start a chat as usual. The cards only fill the box for you — nothing is sent until you press send — and Esc, the × or the “Back” arrow takes you back to the cards.- “Start from an example” (Slides, documents, sheets and charts) — a gallery of six kinds of work: Slides, Document, Spreadsheet, Minutes, Email and Report. Pick one and example requests appear: while the box is empty, hovering an example previews its full wording there; clicking writes it into the box in place of whatever you had typed. For Slides a “Designs” row offers eight designs, three per page (“Previous” / “Next”); the one you pick becomes a chip in the chat box and your request goes out with “Make the slides in the ”…” design.” in front of it. “Import a design” lets you pick
.pptxor.pdffiles, which are attached to the message. “Back” returns to the gallery and clears the design; “Close” returns to the cards and keeps it - “Make work repeatable” (Manuals and workflows for repeat work) — four cards: “Stop re-attaching the same files”, “Turn repeat instructions into a manual”, “Turn routine work into a workflow” and “Run tasks on a schedule”. Each opens a list in place (the first also writes a request about saving your documents into the box) — “Your skills” with “Recommended for you” (skills other members share), “Your flows”, “Scheduled tasks”, or “Ways to use saved documents”. Picking a skill or flow puts it in the chat box as a chip with a line such as “Use the … skill.”; picking a task opens it on the Tasks page; the example cards write a request into the box. “See all” opens the full skills or flows list, with search (and sorting for skills), above the chat box
- “Manage tasks” — opens “Work status” in place. While you are new to Jinba (about your first week) it shows “Getting started with Jinba”: four steps (“Try a recommended use”, “Turn your own work into a skill”, “Once you’re used to it, make a flow”, “Put the flow on a schedule”) whose “Try it” writes a starter request into the box, each marked “Done” once you have a chat, a skill of your own, a connected flow or a scheduled task of your own. The “Switch view” button in the card’s corner flips it to a “Today’s briefing” face; the face you chose is remembered on that device
- “Quick access” — tiles for the agents, skills and flows you pin yourself. “Add to quick access” (the dashed +) opens a picker with All / Agents / Skills / Flows tabs and a search. Pressing a tile shows its details with a “Use” button (tick “Don’t show this again” to skip this step from then on); right-clicking offers “Show details” and “Remove from quick access”. You can also drag a tile onto the chat box — it reads “Drop to use … in the chat” and the drop uses the item right away. Using an agent selects it for the chat; a skill or flow becomes a chip. Pins are kept per workspace in this browser
Screen layout
- Sidebar — The header holds the workspace name with the “Search” button (command palette) beside it; below come New chat, Agents, Flow, Scheduled tasks, the “Configure” section (Skills, Knowledge, Connectors), pinned chats, the “Agents” section, and the “Chats” section of recent ad-hoc chats (grouped into Today / Yesterday / Previous 7 days / Previous 30 days). The “Configure”, “Agents” and “Chats” section headings fold their rows away and open them again; the choice is remembered on that device, a folded section opens on its own when you open a page or chat inside it, and clicking the heading folds it again. A “Manage” link at the end of the Agents heading opens the agent catalog. Clicking an agent row shows or hides that agent’s chats; hovering the row reveals a ”…” menu (“Pin to top” / “Unpin”) and a “New chat” icon that opens the agent’s page. When the sidebar is collapsed to its icon rail, Search is the first icon on the rail
- Message area — The conversation history. The AI’s activity (thinking, running tools, and so on) also appears here. In an agent chat the header at the top reads “<agent name> / <chat title>” — the agent name links to the agent’s page, and the title opens the rename / delete menu
- Composer — Message input, file attachments, and the reference pickers
- Workspace panel — A side panel that previews one file at a time, with the chat’s files, the console log, and — in an agent chat — that agent’s configuration a tab away
Basics
Composing Messages and Attachments
How to send messages, the reference pickers (@ / # / [[ and more), and file attachments
AI Activity and the Workspace Panel
Reading the AI’s progress, interrupting it, and inspecting deliverables and the console log in the preview panel
In-Chat Forms and Suggestions
Flow input forms, multiple-choice questions, AI operation cards, agent setting proposals, and skill suggestions
Managing and Searching Chats
Pinning, renaming, deleting, cross-chat search, and in-chat search
Models and chats
The model used in a chat is decided when the chat is created and stays fixed within that chat. To use a different model, start a new chat — or, when the model behind this chat has changed (the agent’s model, or the workspace default for an ad-hoc chat), switch this chat to the new model (below). You can check the current model on the model badge in the chat screen.When the agent’s settings change mid-chat
An agent’s configuration — system prompt, model, attached skills — is fixed when the chat’s session starts, so editing the agent does not reach a conversation that is already under way. When that happens, a banner appears above the composer: “This agent’s settings were updated — this chat is still running on the previous configuration.” When only the model changed — the agent’s model moved to a newer generation on its own (Notes on choosing a model), once that new configuration has been pushed to the agent, which happens when a new chat is next started with it; or, in a chat with no agent, the workspace’s default model changed — the banner reads “The model has changed — this chat is still running on the previous model.” instead, and its button is “Switch to the new model”. New chats always start on the current settings.- “Switch to the new settings” (or “Switch to the new model”) starts a fresh session on the current configuration; it applies from your next message
- Switching makes the AI forget this conversation’s working context. The visible history stays exactly as it is, but the AI no longer remembers what it had been working on
- You cannot switch while the AI is running, or while the chat is waiting for your answer on a card or an approval — finish or stop that first (“This chat is waiting for an answer, approval, or an operation to finish. Complete the pending step or stop execution before duplicating or switching sessions.”)
When the chat’s agent is archived or un-shared
A chat stays bound to the agent it was started with. If that agent is later archived, or its share with you is withdrawn, the chat can no longer start a fresh session — so instead of failing silently, sending (or trying to switch to the new settings) surfaces the reason and what to do about it:- Archived — “This agent has been archived. Start a new chat with a different agent.”
- No longer shared — “This agent is no longer shared with you. Ask an organization admin to restore access.”
Response language
The AI generally responds in the language you write in. An explicit request such as “answer in English” always takes top priority. The app’s display language (Settings > Personalize > Language) is used as a fallback when the language cannot be determined from your input.When you hit your usage limit
Two limits can temporarily pause a conversation:- Monthly usage — once your organization reaches the usage included in its plan, sending is blocked until the displayed reset date. To raise the limit, contact your organization admin
- This chat’s rolling 24-hour usage — if one chat reaches its own limit, wait for the rolling window to clear or continue in a new chat. Other chats are not blocked by this limit