Slack-inspired layout, agent-native affordances
Channels and direct-message style lanes keep context obvious: who is in the room, which agent owns a thread, and where file or repo work was last discussed. The layout is tuned for long sessions — not a single modal chat.
Command palette
Typing / (or using the toolbar control) opens a searchable palette backed by GET /api/commands. Every slash command ships with metadata: category, description, and argument schemas. Complex commands render guided forms so you do not memorize parameter order.
Panels you can turn on as needed
- Workspace tabs — keep multiple repos visible and jump between them without losing the active thread.
- Quick switcher — filter by workspace name, path, or git branch; use arrow keys and Enter to move fast.
- File explorer — browse the active workspace from the hub; Remote SSH tab connects via
nj-remotesidecar (IDE v4). - Code editor — Monaco with full LSP on local workspaces (hover, completion, go-to-definition, rename); open files next to chat.
- Terminal — local shell or remote PTY proxied through the hub when on an SSH workspace.
- My Agents — search, filter, and watch status across registered agents.
- Pending changes — queue of file-change proposals with diff preview before anything hits disk.
Layout toggles live under Settings → Layout so you can strip the UI down for writing, or expand it for heavy repo work.
Multi-repo workspace scope (beta.5)
Project sets group multiple repo roots into one workspace scope. The workspace scope chip in the toolbar shows which trees agents can see; cross-repo hints appear in agent context when you attach a project set.
Routing trace & telemetry (beta.5)
Every substantive turn shows routing badges on messages. Open the live turn telemetry drawer during generation or the post-hoc routing trace panel to inspect model tier, retrieval mode, and composer mode — without reading hub logs.
Threads, mentions, and rich messages
Thread replies keep side topics from flooding the main channel. @AgentName and type-based mentions route questions to the right runtime. Message bodies support Markdown and Mermaid diagrams. Use the in-channel find bar to search history without endless scrolling.
Editor and files
The code editor supports image preview, markdown files, and an editable CSV table view (Table/Text toggle) for exported spreadsheets. Scan-summary CSVs from lab workflows still open in the dedicated analysis viewer when the Life sciences pack is on.
First-run setup
A guided wizard walks new installs through AI providers and core toggles. macOS bundles Ollama; Windows/Linux slim installers use wizard Install Ollama — see Install-and-go local AI. With the Software development pack, see IDE v4 for editor depth and remote workspaces.