João Moura

Joao Moura is founder and CEO of CrewAI, the open-source multi-agent framework and commercial platform built around role-based crews of agents. He coded from his teens in Brazil, took a detour through law school, then spent years in product engineering before turning agent boilerplate into a company.

Brazil coding and a law-school detour

He started coding around age thirteen in Brazil, later tried law school, then returned to software when product work felt more like home. He became CTO at Palpiteros, then joined Clearbit after flying to San Francisco to meet the team; he rose to senior engineering leadership and later director of AI engineering, building data-activation products and AI experiments remotely from Brazil through Clearbit's acquisition by HubSpot.

From boilerplate pain to CrewAI

While assembling agents for his own engineering work he found too much copy-paste and no clean path from a working prototype to deployment. He published an open-source library to encode role-based crews and tasks; usage spiked, and Oracle was among the first large organizations to reach out asking for help. He left Clearbit after the HubSpot deal, put personal capital into the new company, and drew early backing from Clearbit's founders and HubSpot CTO Dharmesh Shah. CrewAI the company grew around the same crew metaphor: specialized agents, assigned tasks, and orchestration teams can understand.

From prototype crews to production

After the open-source library took off, Moura built a commercial platform around the same crew model so teams could move past demos. The framework stayed how most people try ideas; the product layered on deployment, no-code beside code, guardrails, PII handling, tracing, and the controls companies ask for once agents leave a laptop. In his ODSC AI West 2025 keynote he told the room the industry had over-rotated toward construction: "Building is the tax you pay," with value arriving only once agents run in production. Enterprises, he said, already know how to build agents; they ask for infrastructure, role-based access, versioning, observability, and guardrails — the unglamorous layers that keep systems trusted.

On agency with control

Moura talks about plotting use cases on complexity versus precision, and about hybrid approaches that start with tight control before increasing autonomy as observability improves. He describes CrewAI's events-based mental model as closer to how many engineers already think about queues and messages, and he frames open source as the prototyping layer with commercial value concentrated on production and organizational control. Simplicity, in his telling, is harder than stacking scaffolding.