glitch-rabin

glitch-rabin / swarma

Public

experiment loop for ai agents and swarms

www.swarma.dev
49
9
100% credibility
Found Mar 23, 2026 at 49 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

Swarma is an open-source framework for creating teams of AI agents that run automated experiments to continuously improve their strategies and performance.

How It Works

1
🔍 Discover swarma

You learn about swarma, a simple way to build teams of AI helpers that automatically test ideas and get smarter over time.

2
💻 Get it ready on your laptop

Download swarma and set it up quickly so your AI teams can start working from your everyday computer.

3
👥 Build your first AI team

Choose from ready example teams or create your own group of specialized AI helpers with clear goals and tasks.

4
🚀 Launch the first work cycle

Hit go, and your AI team springs into action, producing results while testing what works best.

5
📊 Watch them score and learn

Swarma automatically grades the output, runs mini-tests, and updates their shared playbook with proven tips.

6
📈 Review progress anytime

Check beautiful summaries of improvements, costs, and evolving strategies to see your team growing.

🎉 Enjoy a self-improving AI swarm

Set it to run daily, and return to smarter, more effective AI helpers that compound their wins effortlessly.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 49 to 49 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 swarma?

Swarma brings closed loop experiment cycles to AI agent swarms in Python, automating the growth playbook: hypothesize via strategy files, execute tasks, score outputs on metrics like content quality, then update playbooks with verdicts. Teams configure via YAML—no code—with flows like researcher -> writer, running via CLI (`swarma init`, `cycle team`, `status`). Pairs OpenRouter routing, SQLite state, and optional QMD knowledge for looping experiments on a laptop.

Why is it gaining traction?

It closes the bh loop experiment gap for agents, evolving static prompts into validated playbooks like growth teams at Uber do, but at swarm scale (50x human speed). Prebuilt squads test hooks, CTAs, SEO—real patterns with program.md guides—plus MCP/REST for Hermes/Claude integration and github experiment tracking via TSV logs.

Who should use this?

AI builders tweaking LangChain experimental github agents for content A/B (hooks vs stories), growth hackers optimizing newsletters/social via experiment loop scrum, indie devs needing self-improving swarms for microcap experiments or defi alpha research.

Verdict

Promising alpha for closed loop experiment needs (49 stars, v0.1.0, multilingual READMEs), but 1.0% credibility score flags early risks—test on non-prod. Strong docs and MIT make it forkable; grab for prototypes till PyPI lands.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.