Hermes Agent v2.0! Huge New Updates: WebUI, Qwen 3.6 Plus FREE, Computer Use, & More!

Key Concepts

• Hermes Agent: An open-source, persistent autonomous AI agent designed for long-term memory, skill reusability, and self-evolution.
• Computer Use: A feature allowing AI to interact with OS-level applications (clicks, typing, scrolling) in the background.
• Multi-Agent Orchestration: A system for managing multiple autonomous agents via a Kanban-style interface.
• Goal Command: A long-running autonomous objective mode that handles planning, execution, review, and error recovery.
• Light Panda: An open-source, machine-optimized browser backend for reliable web automation.
• Qwen 3.6 Plus: A high-performance, long-context (1M tokens) model integrated into the Hermes ecosystem.

---

1. Background Computer Use

Hermes Agent has introduced a native "computer use" feature powered by KUA. Unlike traditional methods that hijack the user's screen or mouse, this implementation operates in the background on macOS (with Windows/Linux support coming soon).

• Functionality: The agent can perform clicks, typing, and app navigation without interrupting the user’s active workflow or changing focus.
• Compatibility: It supports any vision-capable model, including Claude, GPT-4o, Gemini, and local Vision Language Models (VLMs).
• Installation: Users can enable it via the CLI using hermes computer use install or by selecting it interactively through the hermes tools command.

2. Multi-Agent Orchestration and Kanban Integration

Hermes has evolved from a standalone tool into a persistent autonomous workspace.

• Kanban Board: A new visual interface allows users to manage unlimited projects and agent tasks. It tracks the status of operations (To-Do, In Progress, Done).
• Gateway Messenger: Users can subscribe to project updates directly through the configured gateway, allowing for real-time monitoring of agent progress.
• Workflow: By using the hermes update command followed by the initialization and dashboard commands, users can launch a web UI to manage complex, multi-agent environments.

3. The /goal Command

The /goal command introduces a sophisticated orchestration layer for long-horizon tasks.

• Methodology: Instead of a single-turn prompt, the agent enters a continuous loop of planning, executing, reviewing, and retrying.
• Memory Management: It handles hand-offs between different agents, ensuring that subtasks are tracked and completed until the primary objective is met.

4. Infrastructure and Model Integration

• Qwen 3.6 Plus: Available via the News Research portal, this model is highlighted for its 1-million-token context window, making it ideal for complex web development and long-horizon tasks.
• Light Panda: This integrated browser backend replaces traditional, brittle scraping tools. It provides a more reliable, open-source alternative for autonomous web tasks, with automatic Chrome fallback support.
• News Research Portal: Built on OpenRouter, this portal provides access to free models, discounts, and optimized bundles for the Hermes workflow.

5. External Tooling: Tiny Fish

The video highlights Tiny Fish as a solution for the common pain points of web-based AI agents (e.g., broken Playwright scripts or slow search APIs).

• Technical Specs:
  • Search API: Returns results in under 500ms.
  • Browser API: Spins up a stealth Chrome session in the cloud in under 250ms.
  • Data Structure: Converts raw HTML into clean, structured data, which is optimized for agent consumption.
• Value Proposition: It consolidates search, fetch, browser, and agent capabilities under a single API key, with free tiers for search and fetch APIs.

---

Synthesis and Conclusion

Hermes Agent is rapidly positioning itself as a leading open-source autonomous framework. By shifting from simple task execution to a persistent, multi-agent workspace—supported by background computer use, robust browser backends like Light Panda, and long-horizon planning via the /goal command—it offers a highly scalable solution for complex automation. The integration of high-context models like Qwen 3.6 Plus and the ability to manage workflows through a Kanban interface marks a significant step toward reliable, long-term autonomous computing.