The video introduces major updates to the Hermes AI agent, positioning it as a robust platform for multi-agent orchestration, capable of replacing or complementing tools like Open-Crow. The core new features facilitating this are the "Kanban Board" and "Profiles."

The Kanban Board provides a structured way to manage tasks within the Hermes ecosystem. It features six distinct task states: Triage, To-do, Ready, In-Progress, Blocked, and Done. Users can add tasks, assign them to specific "Profiles" (which represent AI agents with distinct personas), set priority levels (0 being the highest), and explicitly define skills the assigned agent should use for task execution (e.g., XURL for web fetching). The board supports detailed task management, including adding comments, defining Parent and Child tasks to establish dependencies and create complex workflows (e.g., an email processing workflow where checking for spam is a parent task to drafting a reply). Tasks can be manually moved between states or automatically transitioned via a "Nudge Dispatcher" function. The system also supports multiple Kanban boards for different projects and allows categorization of tasks (e.g., Marketing, Development). The Web UI provides a comprehensive visual overview and synchronization with other interfaces.

The Profiles feature is fundamental to Hermes's multi-agent capabilities. Each profile defines an AI agent's persona, role, and specific characteristics. These profiles can be created manually or programmatically via other AI agents (e.g., Codex or Hermes itself). The video demonstrates creating multiple specialized profiles (e.g., Project Manager, Grace, Mike, Sunny) to simulate a team of "AI employees."

A significant part of the update is the Slack Integration, enabling users to manage Hermes tasks and interact with agents directly within Slack. A custom plugin, hermes-slackboard, allows access to the Kanban board via the /board slash command. This leverages Slack's Block Kit for interactive UI elements within the chat interface. While drag-and-drop is not supported in Slack, tasks can be moved between states using dedicated move buttons. The setup of this integration is heavily automated, using Hermes agents to configure Slack apps, generate necessary API tokens (App Configuration Token, App-level tokens, Bot tokens), and set up slash commands.

Building on the multi-profile and Slack integration, Hermes introduces a Multi-Agent Meeting feature via a /meeting slash command. This enables simulated virtual meetings where multiple AI agents, each associated with a unique profile and its own Slack app gateway, can participate and contribute based on their defined personas and roles. For example, a "moderator" agent (e.g., Brian) can facilitate the meeting, posing questions and soliciting input from other specialized agents (e.g., Grace for strategy, Mike for technical aspects, Sunny for marketing). The setup for this feature is also highly automated, with Hermes creating profiles, generating Slack app configurations, and setting up independent gateways for each agent to allow simultaneous interaction in a single channel. The system supports various meeting modes (e.g., mixed turns, sequential turns) and can integrate with Text-to-Speech (TTS) services (e.g., Typecast) for voice-based interactions, enhancing the realism of the automated meeting. The meeting concludes with a summary or "recap" provided by the moderator agent, distilling the discussion points and conclusions.

Overall, the updates aim to transform Hermes from a personal AI assistant into a robust platform for orchestrating complex, multi-agent workflows and collaborative tasks, with seamless integration into popular communication platforms like Slack. The emphasis is on automation, personalization, and creating an environment where AI agents can function as specialized, interconnected team members.