Emergence AI Launches AI Agents to Create Other Agents Autonomously

Emergence AI, an AI startup, announced on Tuesday in a blog post that it is upgrading its platform to autonomously create and assemble AI agents. 

This innovation aims to minimise human intervention in enabling AI agents to build and deploy themselves, evolving through recursive self-improvement.

The company stated that the heart of this system is an "orchestrator" capable of coding, planning, spawning tools and agents to handle complex tasks.

The blog post explained that the orchestrator dynamically plans capabilities, reuses existing agents, and creates new ones as needed, then tests and refines these systems through simulation and self-evaluation. This allows agents to learn from failures and optimise their performance, all within human-defined goals.

Today, we’re unveiling the first demonstration of AI agents autonomously creating other agents and multi-agent systems to solve complex enterprise tasks.

⚙️ No manual code
🧠 Built-in planning, memory & verification
🔄 Continuously improving with every task
🤖 Proactively… pic.twitter.com/VMORXTSavs

— emergence.ai (@emergence_ai) April 1, 2025

Furthermore, the orchestrator guides the formation of intelligent architectures without direct human coding.

While the aim remains to achieve an autonomous building process, the company stated that humans remain in the loop to inspect, correct, and guide agent behaviour at any time. 

The company shared an example where the platform automatically generated agents to identify chips with the lowest yield, for a client in the Energy industry, demonstrating its ability to handle complex data analysis.

The platform has also added an intuitive agent SDK supported by a comprehensive agent registry, facilitating seamless integration of second or third-party agents and deployment across various cloud environments.

Emergence AI acknowledges potential challenges, such as bias and goal misalignment, but emphasises that enterprise settings, with structured processes and robust controls, are well-suited to mitigate these risks. The company also stated that development builds on recent advancements in AI, particularly in code generation.

"We believe this shift will result in increasing both the ease-of-use of multi-agent AI and its capabilities,” the company stated. 

Users will have to wait and see if these developments end the hype for AI agents or whether AI agents will be vibe-coded by agents.