foxctl Flow Engine

You orchestrate multi-agent development workflows by executing JSON envelope DAGs with agents, transforms, and skills. Intermediate outputs deterministically route through downstream steps via push-mode output routing, ensuring reliable state and structured report generation.

Push-mode output routing lets agent nodes publish structured JSON back into the engine, automatically triggering downstream nodes. This mechanism turns loosely connected agent steps into an explicit, runnable flow graph with reliable state and routing.

You structure a flow graph using JSON envelope DAGs with node and edge definitions. Agent, transform, and skill nodes consume and produce JSON envelopes, while edges apply transforms and conditions to route intermediate outputs deterministically.

Yes, you can run long-lived multi-agent workflows through a per-workspace daemon. This daemon-backed execution provides caching, agent spawns, and robust logging, status tracking, and error handling for your flow orchestration.

Use JSON DAGs for research pipelines, structured report generation, and automation where intermediate outputs must deterministically route through downstream steps. It solves the challenge of coordinating multi-step AI workflows with reliable state and routing.

Create a flow, add an agent node in push output mode that researches a topic, add a file_write transform node that saves the resulting report to disk, connect them with an edge, and start the flow in your workspace.