EvoMap/evolver is a GEP-powered self-evolving engine for AI agents, addressing the need for auditable, reusable evolution assets in AI development.
Source: README View on GitHub →This project is gaining attention due to its innovative approach to AI agent evolution using Genes, Capsules, and Events, offering a unique solution for creating auditable and reusable evolution assets. Its focus on Genes as a compact representation for agent experience stands out, as evidenced by its performance improvements in scientific code-solving scenarios. The transition to source-available while maintaining a commitment to users also adds to its appeal.
Source: README, Research sectionA protocol for encoding agent experience as Genes and Capsules, enabling auditable and reusable evolution assets. It uses Genes as a compact representation for agent experience, leading to improved performance and robustness.
Source: README, Research sectionA simple and straightforward command-line interface for running evolutions, reviewing changes, and setting up continuous loops. It supports integration with various agent runtimes.
Source: README, CLI Quick Start sectionSupport for integration with major agent runtimes like Cursor, Claude Code, Codex, Kiro, OpenClaw, and opencode, allowing for seamless use within existing workflows.
Source: README, Wire up your agent runtime sectionThe architecture is modular, with clear separation of concerns. It includes an adapter layer for different agent runtimes, a core engine for evolution, and a data storage layer for memory and logs. The use of Genes and Capsules as the core data structure for evolution is a key architectural decision.
Source: Code tree, package.jsonCenter: project; inner ring: core feature modules; outer ring: key dependencies. Auto-generated from core_features and tech_stack.key_deps.
dotenvEvoMap/evolver is suitable for developers and technical decision-makers working on AI agent development. It is useful in scenarios where AI agents need to evolve and adapt based on experience, such as in scientific code-solving, automation, and meta-learning.
Source: READMEv1.80.7 (2026-05-09): Bug fixes and UX improvements.
Source: GitHub ReleasesEvoMap/evolver is a promising project for those interested in AI agent evolution and meta-learning. Its innovative approach to encoding and reusing agent experience through Genes and Capsules offers a unique solution for AI development. Its ease of integration with major agent runtimes makes it a valuable tool for developers working in this field.