Hermes Agent v0.15.0 Released with Major Code Optimization and Multi-Agent Platform Upgrade

icon MarsBit
Share
Share IconShare IconShare IconShare IconShare IconShare IconCopy
AI summary iconSummary

expand icon
NousResearch released Hermes Agent v0.15.0, a major update featuring code optimization and an upgraded multi-agent platform. The run_agent.py file was reduced from 16,003 to 3,821 lines, divided into 14 modules. Cold-start time dropped to 0.8 seconds, and hermes --version now executes in 258 milliseconds. The session_search tool now employs local filtering, achieving a 4,500x speed improvement at no additional cost. The kanban tool now supports full multi-agent collaboration. Security enhancements include defenses against Brainworm attacks. Ecosystem updates integrate Bitwarden and xAI Web Search. A v0.15.1 patch followed to resolve outstanding issues. Altcoins to watch may benefit from such tools as the Fear & Greed Index shifts.

According to Beating Monitor, the NousResearch team has officially released the open-source agent project Hermes Agent v0.15.0. The core run_agent.py codebase has been drastically reduced by 76%, from 16,083 lines to 3,821 lines, and fully restructured into 14 highly cohesive modules, eliminating editor lag during loading. For cold-start and search performance, the development team reduced Termux terminal cold-start time to 0.8 seconds and improved hermes --version speed by 63% to 258 milliseconds by deferring OpenAI dependency loading and optimizing high-frequency function calls. The redesigned session_search tool eliminates large model invocations, replacing them with zero-configuration local filtering, reducing search time to just 20 milliseconds—a 4,500-fold improvement with zero operational cost. The dashboard tool has been upgraded into a full multi-agent collaboration platform: developers can now initialize a Swarm workflow with a single command—hermes kanban swarm—automatically setting up coordinators, parallel workers, independent validators, and synthesis nodes, with added support for single-task model coverage and scheduled startup. On the security front, the new version introduces proactive threat defense against Brainworm attacks, implementing traffic filtering and boundary marking at three critical checkpoints—tool output, memory recall, and external skill loading—to prevent agents from being compromised by malicious files or network services. The ecosystem and third-party integrations have also undergone deep upgrades: the system now supports centralized cloud API key management via Bitwarden, introduces a toolkit enabling single-command loading of multiple skills, and adds an interactive installation panel for the official Nous MCP service directory. Integration with the xAI ecosystem now includes the xAI Web Search plugin, enables Grok-specific execution guidance within the hermes proxy, and allows hermes doctor diagnostics to automatically alert on fine-tuned models with one-click hot migration support. To address critical bugs present at the initial release, the official team rapidly deployed patch version v0.15.1 within hours, completely resolving the infinite 401 reload loop in loopback mode, fixing path resolution failures for bare commands like npx inside Docker containers, and ensuring proper termination of Kanban worker threads under SIGTERM signals.

Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information. Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.