Kimi Code released versions 0.40.0 through 2.1.0 in September, introducing multi-agent collaboration, enhanced UI panels, and improved terminal rendering capabilities.
Agents and Automation
- Tower multi-agent mode: Experimental collaboration mode activated via /tower command or composer menu, with auto-creation of base branches, agent failure tracking, and smaller missions with stricter completion checks in version 2.1.0.
- Updates panel: Shows paginated progress messages from the main agent and sub-agents, available experimentally via environment flag in version 0.42.0.
Interface and Workflow
- Media rail in composer: Web interface displays image and video previews in a media rail within the composer, allowing mentions and keeping previews after queuing or sending.
- TUI fullscreen layout: Experimental fullscreen mode with mouse-interactive UI elements for a smoother terminal experience, added in version 2.1.0.
- Selection comments and quote-to-chat: Now work in diff and per-turn changes panels, and terminal text can be quoted directly into chat or commented on.
- Mermaid diagram rendering: Code blocks render as diagrams in the terminal, with an option to disable via settings or tui.toml.
Platform and Tools
- Plugins panel: Browse and manage plugins from the marketplace in Settings, with support for activating multiple skills from a single message.
- Desktop app command: /desktop slash command (alias /install-desktop) opens the Kimi Code desktop app page in the browser.
- Session management: Command line gains kimi session list command to view your sessions.