HyperAgents

Evolutionary self-improving agent framework powered by LangChain and LangGraph. ⚠️ EXPERIMENTAL

Self-Improving

Uses evolutionary algorithms to improve agent logic (prompts, tools, and TypeScript) and keeps an archive of scored generations with branching lineage.

Pluggable Architecture

Works with multiple LLMs via LangChain and agentic flows via LangGraph; plug in domains with static, LLM judge, or human feedback evaluators.

Agentic Self-Modification

MetaAgent edits the program under evaluation using bash and editor tools; run locally or in Docker for safer sandboxing.

Note: Inspired by HyperAgents (Meta Research, 2026).