explaingit

evomap/evolver

7,407JavaScriptAudience · developerComplexity · 3/5LicenseSetup · easy

TLDR

A CLI tool that helps AI coding agents improve over time by scanning activity logs and generating compact reusable guidance (called Genes) that get passed into future agent runs automatically.

Mindmap

mindmap
  root((evolver))
    What it does
      Scan agent logs
      Generate Genes
      GEP research basis
    Tech Stack
      JavaScript
      Node.js
      npm
    Use Cases
      Cursor integration
      Claude Code hook
      Offline evolution
    Audience
      AI developers
      Agent builders
    Concepts
      EvolutionEvents
      EvoMap platform
      Rollback via git
Click or tap to explore — scroll the page freely

Code map

Detail Auto

An interactive map of this repo's files and how they connect — its source is parsed live in your browser. Click Visualize to build it.

filefunction / class

Things people build with this

USE CASE 1

Automatically improve an AI coding agent without manually tweaking prompts after each session

USE CASE 2

Wire Evolver into Cursor or Claude Code so it fires at session start and feeds Genes back into the agent

USE CASE 3

Track all agent evolution history locally via git and roll back if a Gene makes performance worse

Tech stack

JavaScriptNode.jsnpm

Getting it running

Difficulty · easy Time to first run · 5min

Requires a git-initialized project folder, running outside a git repo will fail with an error.

Was MIT, now GPL-3.0, future versions will be source-available meaning viewable but not freely usable for all purposes. Already-published versions keep their original licenses.

In plain English

Evolver is a command-line tool that helps AI coding agents improve themselves over time by analyzing what they have done and generating structured guidance for future runs. The core idea is that instead of manually tweaking prompts each time an agent makes a mistake, you let Evolver scan the agent's activity logs, identify patterns, and produce a condensed piece of reusable guidance called a Gene. That Gene gets passed back into the agent on the next run, gradually making the agent more effective at the specific work in your project. The approach is based on a research concept called GEP, which the README describes as a way to encode agent experience as compact structures rather than as long skill documents or ad hoc prompt additions. A research paper linked in the README tested this across thousands of controlled trials and found that the Gene representation outperformed skill-document approaches at carrying improvements forward over multiple iterations. Using it starts with one install command via npm. Running evolver inside any git-initialized project folder scans a memory directory for logs, selects an appropriate Gene from its built-in pool, prints a GEP prompt to the terminal, and writes an audit record called an EvolutionEvent. It works fully offline. An optional network connection to the EvoMap platform unlocks skill sharing and leaderboards across different agents. Setup hooks integrate Evolver with agent runtimes like Cursor or Claude Code so it fires automatically at the start of each session or after file edits. All evolution history is stored in a local folder called memory inside your project, and the tool uses git to calculate the scope of changes and to support rollbacks. Running without a git repository will fail with an error message. One licensing note from the README: Evolver started under MIT and later switched to GPL-3.0. Future versions will move to source-available terms, meaning they will be viewable but not freely usable for all purposes. Already-published versions keep their original licenses. The full README is longer than what was shown.

Copy-paste prompts

Prompt 1
Using evolver, how do I initialize a memory directory in my project and generate my first Gene from my agent's activity logs?
Prompt 2
Write a setup hook for Claude Code that runs evolver at the start of each session and passes the GEP prompt back into the agent context.
Prompt 3
I want to wire evolver into Cursor, show me the exact hook configuration so it fires automatically after file edits.
Prompt 4
How do I read and interpret an EvolutionEvent audit record generated by evolver after a run, and how do I use it to improve my next Gene?
Prompt 5
How do I connect evolver to the EvoMap platform to share Genes with other developers using the same agent setup?
Open on GitHub → Explain another repo

← evomap on gitmyhub — every repo by this author, as a profile.

Verify against the repo before relying on details.