Introducing Ebiose: <a href="https://ebiose.com" target="_blank" rel="noopener nofollow">https://ebiose.com</a>
A free, open-source, Darwin-style framework for evolving AI agents that design and improve other agents.
GitHub: <a href="https://github.com/ebiose-ai/ebiose" target="_blank" rel="noopener nofollow">https://github.com/ebiose-ai/ebiose</a>
As developers playing with AI, we’ve all likely struggled with building AI systems that scale or adapt efficiently. Handcrafting agents for every task is slow, rigid, and repetitive. Sounds familiar?
Ebiose, born from a year of R&amp;D at Inria, solves this by letting AI architect agents evolve other agents using survival-of-the-fittest logic. No more manual agent design. No bloated dependencies. Just a lean framework that lives in your codebase.
Why Ebiose?
We’ve burned hours tweaking agent prompts or workflows that don’t generalize. Traditional approaches lack adaptability, and closed systems lock you in.
With Ebiose:
<ul>
<li>Agents run in isolated “forges” with customizable fitness goals.</li>
<li>Architect agents generate and refine candidate agents, improving over time.</li>
<li>Open and flexible, built to be easily stack-agnostic.</li>
<li>Fully open-source, with persistent memory and no forced dependencies.</li>
</ul>
What’s Inside?
<ul>
<li>Evolution engine for iterative agent improvement.</li>
<li>LangGraph-compatible runtime for seamless integration.</li>
<li>Handcrafted architect agent (prompt engineer + graph builder).</li>
<li>Persistent memory per forge for tracking agent progress.</li>
<li>Starter forge examples to kickstart your experiments.</li>
<li>Free cloud runtime credits to run your own forge.</li>
</ul>
Here is an example of a math agent graph created by Ebiose’s math architect agent.
<img src="https://media.daily.dev/image/upload/s--bHb_6uwj--/f_auto/v1751885253/ugc/content_6fe12e15-9ada-4683-be7a-ea2408a2d354?_a=BAMClqZW0" alt="Math agent graph example">
Try a Simple Forge Task
<ol>
<li>Clone the repo: <code>git clone git@github.com:ebiose-ai/ebiose.git</code></li>
<li>Follow the <code>README.md</code> <a href="%5Burl%5D(https://github.com/ebiose-ai/ebiose?tab=readme-ov-file#-quick-start)" target="_blank" rel="noopener nofollow">quickstart</a></li>
<li>Go to <code>/examples</code> and run the <code>math_forge</code> or build your own example.</li>
</ol>
Need Something More Complex?
You need something Ebiose doesn’t support yet. No problem. Join the Discord community discussion, and we’ll work it out.
<ul>
<li>Learn more about Ebiose <a href="https://ebiose.com/blog/ai-for-all-and-by-all" target="_blank" rel="noopener nofollow">here</a>.</li>
<li><a href="https://app.ebiose.com/" target="_blank" rel="noopener nofollow">Create an account</a> and claim free credits to run your own agent forges.</li>
</ul>

The official daily.dev AI community. Led by thought leaders and AI experts. Stay tuned for insights, news and discussions. 

Ebiose is a free, open-source framework that uses evolutionary principles to automatically generate and improve AI agents. Instead of manually crafting agents, it employs architect agents that create and refine candidate agents through survival-of-the-fittest logic. The framework features isolated execution environments called 'forges', LangGraph compatibility, persistent memory, and customizable fitness goals. Developed from research at Inria, it aims to solve scalability and adaptability issues in traditional AI agent development.

Ebiose - open source, agent evolution framework (agents that build agents that evolve)

This entire paradigm breaks down at any significant complexity…
You want <a href="https://github.com/BrainBlend-AI/atomic-agents" target="_blank" rel="noopener nofollow">https://github.com/BrainBlend-AI/atomic-agents</a>