DEVOURED
Kestra 2.0: A new engine, workers anywhere, and flows as agent tools
Kestra 2.0 pivots to a decoupled architecture, enabling distributed worker execution over outbound-only streams and native integration with AI agents.
What: Kestra 2.0 introduces a control plane/data plane split where workers connect via outbound gRPC, eliminating the need for shared database access. It adds native MCP (Model Context Protocol) support to expose workflows as agent tools and features an AI Copilot for inline YAML generation.
Why it matters: By removing the requirement for direct database connectivity, Kestra enables hybrid-cloud and secure private-network deployments that were previously architecturally prohibitive.
Takeaway: Run `kestra-migrate` to identify breaking changes before upgrading from v1.3.x.
Deep dive
- Control Plane/Data Plane Split: Executors and schedulers are centralized, while workers run in isolated networks.
- MCP Integration: Flows are now callable as tools for LLMs via the Model Context Protocol.
- Task Runners: New support for VM-native execution (AWS EC2, Azure VM, GCE) bypassing container constraints.
- Governance: Introduces policy-based rules to validate flow configuration and compliance.
- Loop Task: Replaces
ForEachto prevent executor memory exhaustion by treating iterations as sub-executions. - Storage Cleanup:
PurgeStorageallows for direct cleanup of orphaned execution files.
Decoder
- MCP (Model Context Protocol): An open standard for connecting AI assistants to data and tools.
- Pebble: The templating language used in Kestra for dynamic configuration and expressions.
- ION: A binary serialization format used by Kestra to store execution outputs efficiently.
Original article
Full article content is not available for inline reading.
Explore all 54 articles in today's digest
Fresh dev news, AI-digested daily. Never miss a trend at devoured.fyi