NEW Introducing Apache Airflow® 3: the most significant release in Airflow’s history - Learn More
Support | Log In

Apache Airflow® 3.0

Data Orchestration reimagined for the AI era.

Run tasks anywhere, at any time — all with enhanced security and an intuitive experience that data teams have been demanding.

Try 3.0 on Astro Book a Demo

Airflow 3.0 gives you the freedom to run tasks:

Anywhere
Deploy workers to any environment while maintaining security isolation. Remote Execution lets you run tasks exactly where your data resides — optimizing for cost-efficiency, latency, and data sovereignty without sacrificing centralized control.
Any Time
Unlock new use cases with event-driven scheduling by integrating with message systems like Amazon SQS, so you can trigger your DAGs the moment data arrives – not just on a schedule. By removing the uniqueness constraint for logical dates, users can launch parallel inference runs easily, ideal for GenAI and on-demand ML workloads.
Any Language
Unify your data stack under a single orchestrator with multi-language task execution coming in future 3.x releases. Starting with Golang, Airflow 3 eliminates the need for code rewrites or Python wrappers — choose the right language for each task based on your specific requirements.

Be Enterprise-Ready on Day Zero with Astro

Experience Airflow 3 with exclusive features on Astro, the fully-managed, secure, auto-scaling platform supported by Airflow experts.

Try 3.0 on Astro

Only on Astro

  • Integrated DAG Versioning
  • Remote Execution Agents
  • Astro Executor

3.0 makes Airflow easier
than ever to use.

DAG Versioning

Track every DAG evolution with immutable snapshots tied to each run.

Never wonder which code version produced which outcome - teams can rapidly audit and debug, eliminating confusion and accelerating compliance checks. Troubleshooting now takes minutes instead of hours, giving you clear accountability and pipeline governance that satisfies even the most stringent regulatory requirements.

Learn More

UI Modernization

Navigate easily with an intuitive, modern UI.

Airflow’s UI has been rebuilt from the ground up, with new navigation and views to access to logs and task details quickly, making the experience ideal for both newcomers and power users. This future-focused redesign also opens the door for advanced React plugins and community-driven customizations in Airflow 3.x.

Learn More

Backfills at Scale

Scheduler-driven backfills triggered straight from the UI.

Trigger, monitor, pause, or cancel backfills from the UI or API — no more fragile CLI-based backfills. Large-scale reprocessing jobs run reliably without session timeouts, ensuring consistent performance even for backfills spanning months of historical data.

Learn More

Asset-Centric Syntax

Put your data assets at the center of orchestration.

Define data objects directly with the @asset decorator, simplifying pipeline creation while maintaining visibility into your data's journey. Write less boilerplate code, manage dependencies naturally, and align your pipelines with how you actually think about data.

Learn More

Do more with enhanced security and operational agility.

Task Isolation

Reduce your attack surface by removing direct metadata database connections.

Airflow 3.0 fundamentally enhances security by removing direct metadata database access from worker processes. Instead, tasks communicate through a hardened API that enforces minimal-privilege access and provides clearer audit trails for compliance.

Learn More

Remote Execution

Fulfill data sovereignty and security requirements that regulated industries demand.

Remote workers connect outbound to the orchestration plane, so you can run tasks where they make the most sense—near your data, on specialized hardware, or in compliance-sensitive environments—without sacrificing visibility or control.

Learn More

Run your tasks anywhere at any time.

Event-Driven Scheduling

Move beyond the static cron jobs—orchestrate as events happen.

Event-driven scheduling in Airflow 3.0 integrates with messaging systems like Amazon SQS to start pipelines when data arrives or external events occur. Eliminate polling overhead, reduce latency, and enable near real-time analytics that transform how you conceptualize orchestration timing.

Learn More

AI Inference Execution

Scale AI workloads without the constraints of traditional batch processing.

3.0 removes the single 'execution date' constraint to support multiple parallel inference runs—perfect for GenAI pipelines, large-scale predictions, and on-demand model serving. Now, every AI use case is possible—generative, agentic, predictive or analytical—without operational complexity.

Learn More
Webinar Series: Deep Dive: Apache Airflow 3 with Astronomer

Upcoming Webinars

See Airflow 3 in action.

Join our launch webinar series where Airflow creators break down each new capability in Airflow 3.0, with hands-on implementation strategies and direct access to PMC members.

Reserve Your Spot

Join the next evolution in orchestration.

Try Airflow 3.0 on Astro today and get up to $500 in free credits during your 14-day trial.

Get Started Free Book a Demo

FAQs

Have questions about 3.0? We’re here to help.

When will Airflow 3.0 be generally available?
Apache Airflow 3.0 in now generally available! Astronomer will provide day zero support on Astro, ensuring you can immediately leverage all the new capabilities with enterprise-grade stability and support.
What languages will be supported beyond Python?
Future Airflow 3.x releases will support multiple languages, starting with Golang and eventually expanding to Java and other languages. This allows teams to write tasks in the language that best fits their specific needs without requiring rewrites or workarounds.
Will Airflow 3.0 impact my existing DAGs?
Airflow 3.0 retains backward compatibility for official providers and core functionality. While the underlying architecture has changed significantly, most existing DAGs will continue to work and a few DAGs will require some modifications (case-by-case basis).
How does Remote Execution work with network security policies?
Remote Execution Agents initiate outbound connections only, eliminating the need to open inbound firewall ports. This approach meets stringent security requirements while simplifying network configuration.
Is there a performance overhead with the new architecture?
The new architecture in Airflow 3.0 is designed with performance in mind. Event-driven scheduling and the new backfill integration are asynchronous, preventing resource hogging and improving overall throughput. The decoupled execution model may introduce minimal latency for individual tasks, but this is offset by better throughput, scalability, security, and reliability across your entire orchestration environment.

Build, run, & observe
your data workflows.
All in one place.

Try Astro today and get up to $500 in free credits during your 14-day trial.

Get Started FreeBook a Demo