Atla

About

Find and fix your agent’s most critical failures in hours, not days. Atla helps developers cut time spent on manually reviewing traces. Atla’s LLM judge evaluates your agent step-by-step, uncovers error patterns across runs, and suggests specific fixes—so you know exactly what to fix and why. Atla supports the most popular agent frameworks teams build with, including LangChain, CrewAI, and OpenAI Agents. With real-time monitoring, automated error detection, and prompt experimentation, Atla gives teams the visibility and control needed to confidently ship agentic systems that work. We’re a team of researchers, engineers, entrepreneurs and operational leaders. Our expertise in evals was honed through training our own purpose-built LLM Judges, Selene and Selene Mini, which are available open-source and have been downloaded 60,000+ times.

Founders

• 
  Maurice Burger· Founder

  Co-founder of Atla (S23). Startup veteran @ Syrup, Trim, and Merantix. Masters in CS @ University of Pennsylvania. Half an MBA @ Harvard Business School.

  LinkedIn ↗
• 
  Roman Engeler· Founder

  Co-founder & CTO of atla (S23). AI safety researcher @ MATS. MSc. Robotics @ ETH, Stanford, Imperial.

  LinkedIn ↗

Product launches · 2 launches

• Atla: The improvement engine for AI agents ↗▲ 79Sep 4, 2025

  We help teams building AI agents find and fix recurring failures.

• Selene - The World’s Most Accurate LLM-as-a-Judge ↗▲ 56Feb 26, 2025

  Half of AI’s answers are brilliant, half aren’t. We trained a model to tell them apart.

Change history · none recorded