Conol 0.56.0 — Everything you are working on, in tabs

Until now, a chat in Conol was a panel that slid in beside whatever you happened to have open. That worked when you had one note and one conversation. It stopped working the moment you had four of each. This release rebuilds the main area around tabs: notes and chats are the same kind of thing, they sit in two panes you arrange yourself, and the layout is yours to keep. The Mac app follows the same idea one level up, with as many windows as you want.

Notes and chats live side by side in tabs

Why. The old chat panel had a fixed job: it belonged to the note you were looking at. If you wanted a second conversation you closed the first, and if you wanted to compare two notes you navigated away and lost your place. Everything you were working on had to fit in one document and one panel, so the app kept making you choose.

Tabs remove the choice. A note is a tab, a chat is a tab, and both live in a pane you control. There is no hidden relationship between them any more, which is also why a chat no longer cares whether it started from Home, a workspace or a note. It is just a chat.

How to use it. The main area is split in two: documents open on the left, chats on the right. Each pane has its own row of tabs.

• Open things the way you always did. Clicking a note in the sidebar opens it as a tab on the left; clicking a chat session opens it on the right.
• Rearrange by dragging a tab within its row, or across to the other pane. Drop it anywhere in the target row and it lands where you dropped it.
• Close with the × on the tab, or a middle click. Right-click gives you Close others, Close tabs to the right, Close all, and Move to the other pane.
• Hide the chat side with ⌘J when you want the full width for writing, and press it again to bring it back with your chats still there.
• Cycle tabs with ⌘⌥→ and ⌘⌥←. Close the focused tab with ⌃W in the browser, or ⌘W in the desktop app.

Your layout is remembered per browser tab, not per account. Open Conol in a second browser tab and you get a second, independent arrangement, which is the behaviour you already expect from every other tabbed app. Note that only the active tab in each pane stays mounted, so switching back to a tab reloads its scroll position rather than preserving it.

The Mac app opens as many windows as you want

Why. Tabs solve the problem inside one window. On a Mac with two screens, the next question arrives immediately: put the note on this screen and the chat on that one. The desktop app was built around a single window, so the answer used to be no.

How to use it. Four ways to get another window, all of them equivalent:

• File → New Window, or ⌘⇧N.
• The dock icon, right-clicked.
• The menu-bar icon, under New Window.
• A tab's right-click menu → Move to new window, which hands that exact note or chat to a fresh window and closes the tab here.

Every window keeps its own tabs and its own layout, while sharing the account you are signed in to. Deep links, notification clicks and the ⌃⌘C quick-chat hotkey land in the window you used last. Closing a window closes it for good; closing the last one keeps Conol running in the background the way it always has, so your agents keep working and the menu-bar icon brings the window back.

Copy a tab's identifier, jump back with ⌘K

Why. Sharing a pointer to something in Conol meant copying a URL, and a URL is long, carries your account id, and tells you nothing about what it points at. Worse, it is awkward to hand to an agent mid-conversation or to jot in your own notes.

How to use it. Right-click any tab and choose Copy identifier. You get something short and legible:

note:twjknx6cec20emr4ij4g3oco:c8no55oyvcgh2mefheegaqir
workspace:twjknx6cec20emr4ij4g3oco
chat:hr6ed9gygi2vqfb8zfegpeua
home

Paste one into ⌘K and a Go to row appears at the top with the live title and icon of whatever it names. Press Enter and you are there, in the right pane, with the tab activated if it was already open.

The palette is deliberately forgiving about what you paste. A full Conol link works, with or without the account prefix. A bare id works too: Conol looks it up and tells you whether it is a note, a workspace or a chat. Anything that is not an identifier is treated as an ordinary search, so your normal ⌘K habits are untouched. A chat that has not sent its first message has no identifier yet, so it simply has no menu item.

CoHDL code is highlighted

Why. CoHDL blocks rendered as flat grey text in notes and chat, which made anything longer than a few lines hard to read and easy to mis-copy.

How to use it. Nothing to configure. Paste or write a code block and pick CoHDL as its language; existing blocks that already say cohdl light up on their own. Keywords, unit literals such as 3.3V and 1kohm, attributes, strings and comments all get their own colour, in both the note editor and chat messages, in light and dark themes.

A new chat opens on a real starting page

Why. When chats became tabs, a brand-new chat tab was technically correct and practically bleak: an empty rectangle above the composer, with nothing to tell you what the thing in front of you was for.

How to use it. Press ⌘⇧O, click + on the chat pane's tab row, or hit New chat in the sidebar. You get a greeting and a handful of example prompts. Click one and it drops into the composer for you to edit or send as is. The moment the conversation starts, the transcript takes over.

Fixes

• Typing in the chat box no longer creates notes — the single-letter shortcuts (c, n and the G chords) could fire while you typed in a chat tab, quietly creating a note on the other side of the screen. They now act only when nothing has focus, the way Gmail and Linear handle the same pattern.
• The agent can move notes into a workspace you just created — if you made a workspace mid-conversation, the agent could write new notes into it but every move into it was refused with a scope error. It now checks the access you have at that moment instead of a snapshot taken when the chat began. It also verifies you can actually edit the destination, which it never used to.
• Hiding the chat pane takes one click — the button needed two, because the first click was being undone by the pane it had just closed.
• GLM 5.3 Flash answers at every effort level — the model returned an error at the default effort setting, because its reasoning cannot be switched off and Conol was asking it to.
• Questions from the agent render properly — when the agent asks you something, the explanation above the options now renders its formatting instead of showing raw markup.

One pricing note

Z.ai's launch discount for GLM 5.3 Flash ended on 9 September, so the model returns to its list price: $0.15 per million input tokens and $0.50 per million output tokens, double the promotional rate it launched with in August. Nothing about the model changed, and it remains roughly ten times cheaper than GLM 5.3. Conol passes provider pricing straight through in both directions, which is the same reason GPT-5.6 Sol got cheaper for you last release.