CyrusNuevoDia

CyrusNuevoDia / gepa-research

Public

A plugin for your agentic framework that optimizes code using the GEPA algorithm (Genetic-Pareto LLM-driven search).

autoresearch gepa
17
0
100% credibility
Found Apr 28, 2026 at 17 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

GEPAResearch is an open-source plugin for AI coding assistants that automatically discovers benchmarks in a codebase and uses a genetic evolutionary algorithm to propose and test code optimizations in isolated git worktrees.

How It Works

1
🔍 Discover GEPAResearch

You hear about a helpful tool that lets your AI assistant automatically improve your code by finding what to measure and testing smart changes.

2
🛠️ Add to your AI helper

You easily add the tool to your favorite AI coding assistant, like Claude, with a simple command in the chat.

3
📁 Point it at your project

You tell the tool which part of your codebase to focus on and what kind of test to run to measure improvements.

4
✨ It explores and prepares

The tool scans your project, figures out the best ways to measure success, sets up tests, and runs a starting point — all automatically.

5
🚀 Watch it optimize

You kick off the search, and it safely tries improvements in isolated copies, keeping only the winners that pass your checks.

6
📊 Check the dashboard

A local webpage opens showing the family tree of changes, scores, and progress, so you can follow along easily.

✅ Enjoy better code

Your codebase is improved with higher scores, safe commits you can review or roll back, and a full trail of what worked.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 17 to 17 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is gepa-research?

GEPA-research is a Python plugin for agentic frameworks like Claude Code, Codex, OpenClaw, and Hermes that auto-optimizes your codebase. Point it at a repo: it discovers metrics to improve, instruments benchmarks, runs a baseline, then kicks off GEPA—a genetic-Pareto algorithm from the GEPA research paper that uses LLM-driven reflection to evolve better candidates. Each proposal lands in an isolated git worktree, passes gates like regression tests, and commits winners with full audit trails.

Why is it gaining traction?

It stands out as a Claude agentic plugin (or anthropic agentic plugin) by bridging evals and git for safe, observable optimization—unlike raw Copilot plugins or IDE extensions like IntelliJ agentic plugin. The local dashboard visualizes the Pareto frontier and traces, while GEPA's stall detection and budget caps prevent endless runs. Early adopters love the drop-in skills like `/gepa-research:discover` and `/optimize` for agentic AI in ServiceNow or VSCode agentic plugin workflows.

Who should use this?

AI researchers testing the GEPA algorithm on real codebases, Claude Code power users optimizing evals, or teams with agentic setups in Neovim agentic plugin or GitHub Copilot alternatives. Ideal if you're forking evo-like tools and want git-backed evolution without manual cherry-picking.

Verdict

Grab it if you're in the Claude agentic plugin ecosystem—solid docs and CLI make it usable now, despite 17 stars and 1.0% credibility score signaling early maturity. Run a smoke test on a toy repo first; distributed evals are TODO, but the core loop delivers. (198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.