ChanningLua

ChanningLua / prax-agent

Public

Self-improving agent runtime that learns from experience — test-verify-fix loops, correction detection, cross-project memory, multi-model orchestration.

83
9
100% credibility
Found Apr 15, 2026 at 32 stars 3x -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

Prax is a command-line tool that orchestrates AI agents to repair codebases through automated test-verify-fix loops with persistent memory and multi-model support.

How It Works

1
🔍 Discover Prax

You hear about Prax, a helpful tool that uses smart assistants to automatically find and fix bugs in your code by checking tests and making changes.

2
📥 Get Prax ready

Download and prepare Prax on your computer so it's all set up for your projects.

3
🔗 Link your AI helper

Connect Prax to a smart thinking service so it can understand and work on your code.

4
📂 Open your code folder

Navigate to the folder with your code project where the tests need fixing.

5
💬 Tell Prax what to do

Type a simple instruction like 'run the tests, fix any failures, and stop when they pass' – watch Prax inspect files, make edits, and check again until everything works.

6
🔄 See it in action

Prax shows each step: what it reads, changes it makes, and test results, building memory so next tasks start faster.

✅ Bugs fixed!

Your tests now pass perfectly, code is repaired, and you can continue building with confidence.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 32 to 83 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 prax-agent?

Prax-agent is an open-source coding agent CLI in Python that runs LLM agents on your real codebases, looping through test-verify-fix cycles to automate bug fixes and refactors. Tell it "run pytest, fix failures until tests pass," and it inspects code, edits files, re-runs tests, and persists context across sessions via JSON, SQLite, or vector stores. As a GitHub open source tool, it supports multi-model orchestration with Claude, GPT, or GLM, plus read-only or workspace-write permissions for safe execution.

Why is it gaining traction?

It flips the script on flaky LLM wrappers by prioritizing verification—benchmarks show 10/10 repo repairs in half the time of peers like Hermes. Developers dig the persistent REPL with slash commands (/model, /cost, /todo), making it a practical open source GitHub Copilot alternative for CLI-driven coding AI. Multi-model fallbacks and cost tracking keep runs reliable without vendor lock-in.

Who should use this?

Backend engineers debugging test failures or refactoring auth flows in Python/JS repos. Solo devs analyzing technical debt via natural language queries like "explain login.py and suggest httpx upgrades." Teams needing an open source coding LLM for self-hosted GitHub-like workflows without Copilot subscriptions.

Verdict

Grab it if you're evaluating open source coding agents—strong docs, MIT license, and repo-repair benchmarks punch above its 27 stars and 1.0% credibility score. Still alpha with room for broader lang support, but a solid CLI bet for test-driven AI assistance.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.