What problem does it solve? It lets a .NET developer hand control of a VeloxDev node-graph workflow editor to an LLM agent, so the model can inspect, create, connect, and execute nodes through the same undoable commands the GUI uses, without writing a custom tool layer. ## Core Features & Use Cases - Workflow tool surface: Wire a WorkflowAgentScope onto a tree to give an IChatClient 62 tools for querying topology, mutating the graph, running compiled chains, and reverse-computing node results. - Policy gates and budgets: Gate node execution, generic commands, and interaction handlers behind host-configured flags, plus read/write tool-call budgets and confirmation handlers. - Skills and MCP integration: Load switchable Agent-Skills folders from disk and connect external MCP servers (stdio or HTTP), each usable standalone without the workflow layer. - Use Case: A WPF app developer adds an AI assistant panel that lets users ask the model to build a data-processing pipeline on the canvas, with execution disabled by policy until the host explicitly enables it. ## Quick Start Set up a VeloxDev agent scope on my workflow tree with synchronization context, auto-discovery, and skills, then connect it to my chat client so the AI can inspect and edit the graph.