Virtual Try-On Agent

VTOnAgent is a thin Python facade over the live registry. It is not a LangChain tool loop. For Studio chat, call PlannerAgent / MCP planner_agent.

Default model: kling-ai. A named model in the prompt (Google VTO, FASHN, FLUX VTO, Segmind, Qwen-Image, …) pins that VTON registry id.

from tryon.agents.vton import VTOnAgent

agent = VTOnAgent()
result = agent.generate(
    person_image="person.jpg",
    garment_image="shirt.jpg",
    prompt="Use FASHN",
    dry_run=True,
)
print(result["service"], result["model"], result.get("call"))
# vton  fashn-tryon-max  ...

Live path: run_recipe("vton", …) → invoke_model("vton", model, …) — the same runner as MCP vton_kling_ai.

See Planner Agent for chat, filtering, and MCP.