A cross-platform orchestration application that uses DAG workflows to coordinate multiple AI agents.

DAG多智能体编排跨平台应用工作流图智能体协作任务依赖

Hyacinth is a DAG-based cross-platform multi-agent orchestration application. Instead of treating agents as parallel chat participants only, it models agent work as directed workflow nodes with explicit dependencies, execution order, and reusable orchestration patterns.

ModelDAG

RoleOrchestrator

AgentsMulti-agent

TargetCross-platform

Repository

https://github.com/ChivaLryCieux

Technical Stack

Tauri 2DAGMulti-agent orchestrationCross-platform appWorkflow graphAgent routingTask coordination

01 / DAG workflow

Agent work is represented as a directed graph with clear dependencies.

Hyacinth uses DAG structure to make multi-agent coordination explicit. Each node can represent an agent task, transformation, or decision step, while graph edges define how outputs move through the workflow.

02 / Multi-agent orchestration

Multiple agents can be composed into repeatable execution patterns.

The application positions agents as coordinated workers rather than isolated chat windows. A workflow can route context between agents, sequence specialized responsibilities, and make collaboration between models easier to inspect and reuse.

03 / Cross-platform runtime

The orchestration surface is designed to travel across desktop and mobile environments.

Hyacinth is positioned as a cross-platform control layer for agent workflows, allowing orchestration logic to remain consistent while the interface adapts across supported device targets.

What I Built

Uses DAG structure to model agent dependencies, execution order, and data flow.

Coordinates multiple AI agents as a workflow instead of a loose group chat.

Supports reusable orchestration patterns for complex AI-assisted tasks.

Keeps agent responsibilities and handoffs visible through graph-based composition.

Targets cross-platform use so the same orchestration logic can run across device contexts.