2.0.0

Agent Console 2.0 is a milestone release. Save the prompts you reuse and fire them in one click, run many agents in parallel without losing work to an accidental close or restart, find anything you’ve done, and drive it all from the keyboard – without leaving your vault.

What’s new

Quick Prompts – your kickoffs, one click away

• Save the prompts you type over and over – turn the kickoffs you retype every session ("debrief this meeting", "summarize the selection", "get the latest on this project") into little notes, then fire them without retyping. Launch a prompt from a chip above the composer, a ! search right in the message box, or the command palette. Context-aware chips appear only on the notes they fit, so the right kickoff is one tap away and nothing clutters the rest. Prompts can pull in your current selection, open in a fresh tab, and are just notes – so you can version, grep, and share them like anything else in your vault.

Your sessions are safe

• Reopen the panel and your chats come back – close the Agent Console panel and open it again in the same session, and your tabs re-adopt exactly as they were. Combined with restart persistence, there’s no longer a way to lose a panel full of live chats by accident.
• Undo a closed chat – closed the wrong tab? Bring it back with its full transcript, agent, label, and pinned notes, the same way you’d reopen a closed browser tab.
• A safety check before you close everything – closing the panel with two or more chats open now asks first, so one keystroke can’t drop a room full of agents.
• Your half-typed message survives – unsent text in the composer is kept per tab across a panel close or an Obsidian restart, so you never lose a prompt you were still writing.

Find your past work

• Search your whole chat history – find that conversation you half-remember by title or by what was actually said inside it, then reopen any session in a tab and pick up where you left off.
• Take a conversation to another agent – switch a chat from one agent to another and bring the conversation along, shown to the new agent as a read-only block you can collapse. Nothing is ever cleared without asking first.
• Tabs name themselves – each tab picks up a short title from what you’re working on, so a wall of tabs stays readable – and you can still rename any one yourself.

Drive it all from the keyboard

• Every control is keyboard-operable – tab to any button and see a clear focus ring, open menus and dropdowns with Enter or Space, and dismiss them with Escape. Nothing in Agent Console needs a mouse.
• Your hands stay on the keys – bind hotkeys to switch tabs, show the tab list, and fire quick prompts; and after you pick a model, mode, or setting, focus returns to the composer so you keep typing without a click round-trip.

Shape how agents behave

• Agents know they’re in Obsidian – from the first message, each agent gets a short note about its surroundings: that it’s in Obsidian, that your replies render links, math, and diagrams, and that it can work with your notes. You choose exactly what it’s told.
• A calmer settings pane – settings are grouped by what you’re trying to do, and each agent’s options fold away until you need them.
• A working directory per chat – point each chat, or a specific agent, at the folder it should work in, and that choice sticks across restarts.
• Your first chat just works – on a fresh install, Agent Console picks an agent you already have, makes it the default, and opens the panel for you.

Stay on top of parallel work

• Line up your next message while the agent works – type your follow-up and press Enter; it queues and sends the moment the current reply finishes, so you never wait on the cursor.
• Know which agent just finished – the "agent finished" notification names the tab it came from, and clicking it jumps straight to that tab.
• Every shared link, one click away – notes and web pages the agent points you at collect in a list in the chat header, with files it just created this session set apart. Jump back to anything without scrolling.

Bug fixes

• The chat stays put while you type – the view no longer jumps to the bottom when you’re writing a long message in the composer (#194, #162).
• Your model choice sticks for Kiro CLI – the model picker and header model name are back after the agent-protocol upgrade briefly dropped them (#151, #155).
• Keyboard-opened menus land where you expect – header menus anchor to their button, toolbar dropdowns show a focus ring, and Escape returns focus cleanly (#122, #123, #124, #140).
• Mentions search the way you’d expect – multi-word @ mentions match, and Escape or Shift+Enter dismisses the mention popover (#171).

Internal changes

• Migrated to the Agent Client Protocol SDK 1.0 (#131).
• Added a jsx-a11y lint guard so new interactive UI stays keyboard-accessible, enforced on every change.
• Centralized all network calls behind a fixed host allowlist with a test that fails any PR adding a call outside it (#85).
• Extracted pure resolvers for read and gating decisions, and routed every saved-session write through one shared store so titles and history can’t race.
• 1,800+ automated tests across the plugin’s core logic, plus a token-efficiency benchmark and a performance baseline checked on every change.

Install

Settings → Community plugins → Browse → "Agent Console" → Install → Enable. Existing users update via Settings → Community plugins → Check for updates.