CrewAI OSS is going GA

During our first year at CrewAI, we set ourselves a goal.

Build the framework capable of orchestrating a billion autonomous agents — securely, reliably, at scale.

Little over a year later, we’ve done more than that:

Today we’re announcing CrewAI OSS v1.0, the same open-source core now powering 1.4 billion Agentic Automations across the world’s largest enterprises, on all degrees of complexity.

Two years ago we bet that large-language models shouldn’t be micromanaged—they should be delegated real work.

Last week, Andrew Ng called us visionaries for betting early on one idea:

As models get better, engineers will remove the scaffolding [all rules and forced structure] and delegate more, not less.

Andrew Ng invested in CrewAI over a year ago — joined by Dharmesh Shah (CTO HubSpot) and Amjad Masad (CEO Replit) — because they saw what was coming: developers moving from chatbots to crews, from scripts to flows, from prototypes to production agents.

LLMs alone aren’t the finish line — they’re the foundation. CrewAI exists to bridge the gap between imitation and intelligence, turning raw model capability into coordinated reasoning and memory through multi-agent orchestration.

And so did we. So we went to work!

• 1.4 B+ Agentic Executions (450 M+ / month and rising)
• 60 % of the Fortune 500 running Crews & Flows
• 40 k GitHub stars, 5.2 k forks, 1.8 M Downloads / Month
• 115 OSS Releases and events in 30 Cities Worldwide

Built by the Crew, for the Crew

CrewAI grew through builders.
LLMs might have their limits, but builders haven’t hit theirs.

Thousands of developers are already filling the gap—adding memory, reasoning, and collaboration through CrewAI.

That’s how this all started: one open repo, a few obsessed contributors, and a shared hunch that agents working together could do what single models never will.
Fast-forward a few months and that hunch turned into a movement.

Developers who forked our repos, filed issues at 3 a.m., shared templates, and hacked together the first Agentic workflows before there was even a name for them.

Today, that community spans every time zone and company size — a global crew building the future of agentic software.

Every fix, feature, and improvement in v1.0 traces back to someone in that network — built in the open, improved by thousands, scaled by everyone.

Proven in Production

Every day, thousands of Crews and Flows are running inside production systems across finance, consulting, tech, manufacturing, and logistics.

Some of these companies have thousands of engineers working with CrewAI, major companies like IBM, Microsoft, Procter & Gamble, Walmart, SAP, Adobe, PayPal and many more.

From generating documentation to reviewing code, triaging incidents, automating RevOps processes, Know Your Customer automations, summarizing meetings, and automating customer ops — agents built on CrewAI are already working quietly in the background of the world’s largest companies.

LLMs can mimic tasks impressively — but they can’t deliver end-to-end workflows. That’s where Crews deliver

With v1.0, the same production-grade framework these enterprises use is now stable, open, and ready for anyone — from startups automating support tickets to global teams orchestrating hundreds of agents a day.

What’s New in v1.0

CrewAI 1.0 is everything we’ve learned from thousands of builders, tens of millions of runs, and months of obsessive iteration.

Built from the ground up for complexity, not just simple automation. Because real agent systems don’t follow linear chains—they evolve.

Major highlights:

• More Support for Complex Systems
  • Flows now support massive complexity — a thin, low-level layer with full control for advanced orchestration that is easy to use.
• Stable Crew & Flow APIs
  • Predictable, documented, and versioned. The abstractions are now locked for long-term use.
• Native, Free Tracing
  • Every execution now ships with visibility. No setup, no cost — just instant observability for every run so it's extremely easy to debug, no third-party tool required. Because scale reveals what prototypes hide.
• Unified CLI
  • The same tool for building locally, testing in staging, and deploying to production. One command, one flow.
• More Integrations
  • Improved support to MCPs and integrations with CrewAI AMP.
• Deterministic Runs & Better Logging
  • Easier debugging, cleaner output, and consistency across runs — addressing the top community request. Ability to user training, memory, testing, and hooks and Flows to have all control necessary for more deterministic runs.

Why Agents Matter Now

LLMs are incredible — but they start from zero every time. They don’t remember, plan, or collaborate.

That’s part of the gap CrewAI fills.

Agents give models structure, memory, and agency — turning raw intelligence into systems that can think, act, and learn together.

As Karpathy said this week, LLMs are still brilliant mimics, not minds.

Crews and Flows are how we start bridging that divide.

• With Crews, we give language models defined roles and goals — specialists that can reason, debate, and self-correct.
• With Flows, we connect those roles into coordinated chains of work — context-aware, event-driven, and observable.

This isn’t about bigger models. It’s about better orchestration — making the intelligence we already have useful, reliable, and aligned with human goals.

CrewAI isn’t trying to make LLMs sentient. We’re making them effective, v1.0 gives developers the tools to build systems that do

Enterprise and Partners

CrewAI started as an open framework — today it powers many use cases on enterprises across the globe.

As teams scale, they graduate to CrewAI AMP — our platform built for ease of use, trust, governance and control.

Through AMP, organizations gain fine-grained RBAC, secure deployments, audit trails, and full telemetry — the same foundations used by our largest customers.

We’re proud to collaborate with IBM, PwC, and NVIDIA on enterprise-grade agent orchestration — and partner with ecosystem leaders like Arize, Databricks, Galileo, HPE, and many more to bring CrewAI to new platforms and verticals.

CrewAI is becoming the connective tissue of the enterprise agentic stack — open at its core, scalable by design.

What’s Next

CrewAI v1.0 is just the foundation.

Over the next few weeks we’re releasing new features, working on entire new breakthrough improvements, expanding the company even further and getting as many use cases live as we can.

You can learn all about it on our new conference: CrewAI Signal and learn from the some of the major people driving this industry like

• Andrew Ng (CEO & Founder at DeeplearningAI)
• Aaron Levi (CEO & Founder of Box)
• Amjad Massad (CEO & Founder at Replit)
• Some of our customers, our team and our community!

Together with the community, we’re on our way to orchestrating the next 10 Billion Agentic Automations.