Self-Training Agents: Hermes Agent, HF Traces, Skills, MCP & Finetuning — Merve Noyan, Hugging Face

2026-05-13 · Source: AI Engineer · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Cloud Computing & IT Infrastructure · Depth: Intermediate, long

Summary

Hugging Face is fostering an "Open Agent Ecosystem" by providing infrastructure and tools that enable the development and deployment of AI agents using open-source models. The platform emphasizes the benefits of open source, including privacy, cost-effectiveness, and performance parity with closed models, citing examples like GLM 5.1. Hugging Face Hub, which hosts nearly 3 million models, datasets, and spaces, facilitates this ecosystem. Key features include filtering for agentic models, benchmark datasets like SWE Bench Pro for model comparison, and inference providers for optimized model deployment. The Hub also offers tools for local coding agents (e.g., Llama Agent, Pie), the Hermes agent for enhanced memory management, and Hugging Face Traces for hosting and analyzing agent session data. New "skills" allow agents to manage repositories, run jobs, and even train models on remote or local infrastructure, automating complex tasks like instance sizing and cost calculation.

Key takeaway

For AI Engineers building and deploying agentic workflows, Hugging Face Hub offers a robust open-source foundation. You should explore the platform's agentic model filters, benchmark datasets, and specialized tools like Hermes agent and Hugging Face skills to streamline development, ensure privacy, and optimize deployment costs. Leverage the automated training and infrastructure management capabilities to accelerate your agent projects and reduce manual overhead.

Key insights

Hugging Face provides a comprehensive open-source ecosystem for developing, deploying, and managing AI agents and models.

Principles

• Open source models offer privacy and cost benefits.
• Open models now rival closed models in performance.
• Vision capabilities are becoming standard for LLMs.

Method

Hugging Face skills enable agents to automate tasks like model training, dataset exploration, and demo building by interacting with the Hub's infrastructure and APIs, handling backend complexities.

In practice

• Use GLM 5.1 for strong open-source agent performance.
• Deploy Hermes agent for advanced memory management.
• Utilize Hugging Face skills for automated model training.

Topics

• Open Agent Ecosystem
• Hugging Face Hub
• Hugging Face Skills
• Hermes Agent
• Vision LLMs

Best for: AI Engineer, Machine Learning Engineer, MLOps Engineer

Related on AIssential

• OpenAI Agents SDK Crash Course (with Hugging Face Models) — OpenAI Agents SDK simplifies AI agent creation with minimalist primitives, supporting both OpenAI and Hugging Face models.
• Hermes Agent - Complete Local Tutorial | Private Setup with Gemma 4 & llama.cpp | Better OpenClaw? — Hermes Agent provides a local, open-source agentic harness for LLMs, enabling advanced tool use and self-improvement.
• The PR you would have opened yourself — A new Skill and test harness streamline porting transformer models to MLX, addressing agent-generated PR quality.
• Train AI models with Unsloth and Hugging Face Jobs for FREE — Unsloth and Hugging Face Jobs enable efficient, cost-effective LLM fine-tuning, automatable via coding agents.
• Hermes Agent Foundation Update: Real-Time Agents, DeepSeek V4 FREE, Native Windows Support, & More! — Hermes Agent's 0.14.0 update unifies AI tooling, boosts performance, and expands autonomous capabilities.
• Automating LLM Post-Training with Hugging Face’s ml-intern — Hugging Face's ml-intern models ML research as a multi-step workflow, not just a tool-calling chat assistant.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI Engineer.