Self-Evolving AI Agent Recommendations

After Hermes, these self-growing Agents are the most worth pursuing!
Brothers, after Hermes Agent became a sensation online, everyone is asking: besides Nous Research, this increasingly smarter digital clone, are there similar self-evolving AI Agents?
The answer is: Yes, and a wave of self-evolution has already exploded in 2026.
These projects all capture the essence of Hermes: closed-loop learning, autonomous skill refinement, persistent memory, self-optimization—no longer just temporary workers, but truly long-term partners that grow. I’ve browsed GitHub, arXiv, Reddit, Zhihu, Bilibili across the web, and selected the top 5 open-source/framework projects closest to Hermes, ranked by similarity + practicality. Each includes GitHub links, core features, ease of onboarding, and comparison with Hermes to help you quickly choose:
🟢
1. EvoAgentX (Most recommended! The most similar to Hermes’ “Evolutionary Ecosystem”)
• GitHub:
• Core features: Complete self-evolution framework supporting build→evaluation→evolution closed loop. Uses retrieval augmentation, mutation, guided search, and other strategies to automatically optimize agent workflows and skills. Built-in Self-Evolution Engine, allowing the agent to “self-upgrade” like software iteration. Also accompanied by research paper “Self-Evolving AI Agents”.
• Comparison with Hermes: Hermes emphasizes personal persistent memory + skill files, while EvoAgentX excels in automatic evolution of multi-agent workflows, suitable for complex tasks. Both can “get stronger with use,” but EvoAgentX’s evolution is more automated (no manual patches needed).
• Ease of onboarding: Intermediate (supports Ollama, local deployment).
• Suitable for: Developers or power users interested in “agent ecosystem” evolution.
Hermes is a personal clone; EvoAgentX is an evolving legion. One of the hottest self-evolution frameworks in 2026.
🟢🟢
2. aiwaves-cn/agents (Agents 2.0) (Data-driven self-evolution king)
• GitHub:
• Core features: Symbolic learning + data centralization for true self-evolution. Agents automatically extract experience from interactions and update their logic, forming a “lifelong learning” closed loop. The paper title is “Symbolic Learning Enables Self-Evolving Agents”. Supports autonomous language agents, with memory and skills growing with tasks.
• Comparison with Hermes: Both emphasize “learning from experience,” but aiwaves is more academic + data-driven, with skill evolution leaning toward symbolic (high interpretability). Hermes’ four-layer memory system is more practical; this one’s lifelong learning is more hardcore.
• Ease of onboarding: Intermediate (Python framework).
• Suitable for: Researchers/developers seeking explainability and self-growth.
If Hermes gets smarter the more you use it, this one gets smarter + traceable the more you use it.
🟢🟢🟢
3. Agent0 Series (aiming-lab/Agent0) (Zero-data self-evolution dark horse)
• GitHub:
• Core features: Zero-data self-evolution—no human-labeled data needed, directly evolving from tasks through integrated reasoning and self-iteration. Supports Agent0 (language agent) and Agent0-VL (visual-language agent), emphasizing “Tool-Integrated Reasoning” for autonomous growth.
• Comparison with Hermes: Hermes relies on user interaction + skill files to accumulate experience; this is more aggressive—“zero-data start” can self-evolve from scratch. Both support persistent learning, but Agent0 is better suited for exploratory tasks.
• Ease of onboarding: Intermediate-Advanced.
• Suitable for: Tech enthusiasts wanting to experiment with “pure autonomous evolution.”
Hermes is companionship-based evolution; this is a self-reliant evolution from zero to hero.
🟢🟢🟢🟢
4. Tencent/SelfEvolvingAgent (WebEvolver etc.) (Produced by Tencent AI Lab)
• GitHub:
• Core features: WebEvolver! Webpage agent self-improvement + co-evolving world model. Agents continuously optimize themselves in real environments, building dynamic world models to achieve self-evolution in long-term tasks.
• Comparison with Hermes: Hermes is versatile + multi-platform, while Tencent’s focus on webpage/environment interaction for self-evolution, with world models making it stronger in dynamic scenarios. Enterprise backing with high code quality.
• Ease of onboarding: Intermediate.
• Suitable for: Those working on browser automation, web agents.
Hermes is an all-round evolution; this is a specialized evolution for web battlefield.
🟢🟢🟢🟢🟢
5. CharlesQ9/Self-Evolving-Agents + Awesome List (Research + tool collection)
• GitHub: +
• Core features: Not a single framework, but a panoramic survey of self-evolving agents, including memory evolution, reflection capabilities, lifelong learning, etc. The Awesome list compiles all top papers, benchmarks, and open-source projects from 2025-2026, providing a one-stop understanding of the ecosystem.
• Comparison with Hermes: Hermes is practical; this series is “theory + toolbox.” For those wanting an in-depth understanding of self-evolution mechanisms, a must-see!
• Ease of onboarding: Low (research-oriented) → Medium (playing with code).
Hermes is a solo soldier; this is a full self-evolution toolkit + roadmap.