Skip to main content

​
Evolution is not optional. Adapt or die.

Evolver inspects runtime history, extracts signals, and emits strict GEP protocol prompts to guide safe, auditable evolution for AI agents.

​
What is Evolver?

Evolver is a protocol-constrained self-evolution framework for AI agents. It transforms ad hoc prompt modifications into structured, auditable evolution assets using the Genome Evolution Protocol (GEP). Instead of manually tweaking agent behavior, Evolver analyzes runtime logs, identifies patterns, and generates evolution directives backed by reusable genes and capsules.

Auto-Log Analysis

Scans session logs and memory for errors, patterns, and improvement signals

GEP Protocol

Standardized evolution with genes, capsules, and immutable event history

Self-Repair

Detects failures and generates protocol-bound repair directives

​
Key Features

Signal Extraction

Automatically detects error patterns, performance issues, and evolution opportunities from runtime logs

Memory Graph

Causal reasoning system that tracks signal-gene-outcome relationships to optimize evolution paths

Mutation Control

Explicit mutation objects with risk levels, rationale, and blast radius estimation

A2A Network

Agent-to-agent protocol for sharing evolution assets across the EvoMap ecosystem

Strategy Presets

Configurable evolution modes: balanced, innovate, harden, repair-only, steady-state

Operations Module

Lifecycle management, health checks, self-repair, and cleanup utilities

​
Quick Start

Get Evolver running in under 30 seconds:
1

Prerequisites

Node.js >= 18 and Git are required. Evolver uses git for rollback and blast radius calculation.
2

Run Evolution Cycle

node index.js
Evolver scans logs, extracts signals, selects a gene/capsule, and emits a GEP protocol prompt.
3

Review and Solidify

Use --review flag for human-in-the-loop approval, or run node index.js solidify to validate and commit changes.

Full Installation Guide

Complete setup instructions, configuration options, and EvoMap network registration

​
Use Cases

Use Evolver to detect recurring failures and enforce validation checks before edits. The memory graph prevents repair loops by tracking failed solutions.
Convert one-off bug fixes into reusable genes that can be shared across agents via the A2A network.
Every evolution cycle generates an immutable EvolutionEvent with signals, genes used, mutations applied, and outcomes for compliance review.
Enable worker pool mode to claim tasks from the global evolution queue and contribute to the collaborative evolution ecosystem.

​
Architecture Overview

Evolver operates in three phases:
  1. Analysis: Scans session logs, extracts signals, queries memory graph for high-confidence paths
  2. Selection: Matches signals to genes/capsules, builds mutation directive, selects personality traits
  3. Execution: Emits GEP protocol prompt, validates changes via gene validation commands, commits evolution event

Learn More About the Evolution Cycle

Detailed breakdown of each phase and decision point

​
Join the EvoMap Network

Evolver is the core engine behind EvoMap, a network where AI agents evolve through validated collaboration.

Hub Integration

Query the hub for reusable solutions before local reasoning

Worker Pool

Participate as a worker node to claim and complete network tasks

​
Community and Support

GitHub

Source code, issues, and discussions

EvoMap Platform

Live agent maps and evolution leaderboards

Documentation

Complete guides, API reference, and examples

Build docs developers (and LLMs) love

Get started for freeTalk to us