EvoMap

evolver

// summary

Evolver is a GEP-powered self-evolution engine designed to transform ad hoc AI agent prompts into auditable and reusable evolution assets. It scans runtime logs to identify patterns and emits protocol-bound prompts that guide agents through structured self-repair and optimization cycles. The system supports various host runtimes and offers optional network features for collaborative skill sharing and decentralized validation.

// technical analysis

Evolver is a GEP-powered self-evolution engine designed to transform ad-hoc AI agent prompt adjustments into auditable, reusable assets. By scanning runtime logs and error patterns, it generates protocol-bound prompts that guide agents through structured evolution cycles, effectively turning isolated tweaks into a managed intelligence system. The project prioritizes security and traceability through a strict validation model, ensuring that autonomous agent changes remain within defined safety boundaries while providing a robust framework for continuous improvement.

// key highlights

Automated log analysis scans memory and history files to identify recurring errors and patterns for targeted agent improvement.

The GEP (Genome Evolution Protocol) provides a standardized, auditable framework for managing reusable evolution assets like genes and capsules.

Configurable strategy presets allow users to balance evolution between innovation, optimization, and stability based on current system needs.

A built-in safety model restricts validation commands to a whitelist and prevents arbitrary shell execution, ensuring secure autonomous operations.

The Skill Store enables users to download and share reusable agent skills, fostering collaborative intelligence across the EvoMap network.

Integration hooks for platforms like Cursor and Claude Code allow Evolver to seamlessly influence agent runtimes via standardized stdout directives.

// use cases

Hardening flaky agent loops by enforcing validation before applying edits

Encoding recurring fixes as reusable Genes and Capsules for consistent agent behavior

Generating auditable evolution events to ensure traceability and compliance in agent workflows

// getting started

To begin, ensure you have Node.js 18+ and Git installed, then install the CLI globally using 'npm install -g @evomap/evolver'. Once installed, you can run 'node index.js' to perform a single evolution cycle or use '--loop' to run it as a background daemon. For platform-specific integration, use the 'evolver setup-hooks' command to wire the engine into your preferred agent runtime.