AI-Powered Airflow Management: From Desktop Assistant to Embedded UI Plugin
Description
Explore the evolution of AI-powered Apache Airflow management through two complementary approaches: enhanced desktop capabilities and native UI integration.
Part 1: Airflow MCP Enhancements
Discover powerful new features for Claude Desktop users:
Safe Mode by Default: Read-only operations protect production environments, with opt-in write capabilities
Smart DAG State Management: Pause all DAGs before outages and automatically restore only previously-active ones
Predictive Scheduling: Forecast when DAGs will run, even complex event-driven workflows with no fixed schedule
Part 2: Introducing airflow-chat Plugin
For organizations unable to use Claude Desktop due to data sharing, compliance, or policy concerns, we've brought the same conversational interface directly into the Airflow UI:
Flexible Deployment: Run embedded in the webserver, as separate services, or in fully decoupled architecture
Multi-LLM Support: Choose between Anthropic, Amazon Bedrock, or OpenAI backends
Complete Control: Keep everything within your infrastructure and compliance boundaries
Built-in Access Controls: Admin and role-based permissions out of the box
Whether you prefer desktop integration or need an embedded solution, these tools transform how you interact with Airflow, using natural language to inspect DAGs, manage states, and predict pipeline behavior with just a few prompts.
Ideal for: Data Engineers, DataOps Engineers, Platform Engineers, and teams managing Apache Airflow across development and production environments.
Speaker
Egor Tarasenko
Data & AI Engineer, Ponder
Egor is an experienced Data & AI Engineer with over five years of industry experience. Currently, he works at Ponder, where he helps deliver data and AI services tailored for educational institutions. His passion lies in building practical, impactful solutions that make data more accessible and valuable.
He's also deeply engaged in the open source community through their work at Ponder. This includes developing AI and data libraries, creating plugins and integrations for tools such as Superset, Airflow, dbt, and others, as well as building innovative solutions for technologies like Trino and Apache Iceberg.