AI agents
Drop GarmentVerse into Claude, GPT, Cursor, or any agent — discoverable via OpenAPI 3.1 and a forthcoming MCP manifest.
Bring your API key, point the agent at the spec, ship. No special SDK wrapping needed — every endpoint is in the public OpenAPI spec at https://garmentverse-api-807711804803.asia-south1.run.app/openapi.json.
#Use as a Claude tool
Define a tool with the request schema and dispatch try-on calls inside your tool handler.
import Anthropic from "@anthropic-ai/sdk";
import { createClient } from "@garmentverse/sdk";
const anthropic = new Anthropic();
const gv = createClient({
baseUrl: "https://garmentverse-api-807711804803.asia-south1.run.app",
apiKey: process.env.GARMENTVERSE_API_KEY!,
});
const tools = [{
name: "tryon",
description: "Try a garment on a person. Returns a job id; poll until succeeded.",
input_schema: {
type: "object",
properties: {
person_image: { type: "string", description: "GarmentVerse file_id" },
garment_image: { type: "string", description: "GarmentVerse file_id" },
category: { type: "string", enum: [
"saree", "lehenga", "kurta", "sherwani",
"dress", "tshirt", "suit", "other"
]},
subtype: { type: "string", description: "e.g. nivi, bengali" },
views: {
type: "array",
items: { type: "string", enum: [
"front","three_quarter_left","side","three_quarter_right","back"
]},
},
},
required: ["person_image","garment_image","category"],
},
}];
// Inside your tool dispatcher:
const job = await gv.images.tryon({
mode: "garment_to_person",
person_image: input.person_image,
garment_image: input.garment_image,
garment: { category: input.category, subtype: input.subtype },
views: input.views ?? ["front"],
consent_token: true,
});#Use as a GPT Action
Custom GPTs and ChatGPT Actions can import the OpenAPI spec directly. Configure the action with Bearer auth and your API key.
OpenAPI URL: https://garmentverse-api-807711804803.asia-south1.run.app/openapi.json
Authentication: Bearer
Paste your sk_live_… key from the dashboard#Use over MCP
A Model Context Protocol manifest at /.well-known/mcp.json exposes resources (workspaces, files, jobs) and tools (tryon, catalog, video) to MCP-compatible clients — Claude Desktop, Cursor, Goose, Zed.
#Cursor / Continue / Aider
Any IDE agent that supports OpenAPI tool import works the same way: drop in the spec URL and your API key.
{
"mcpServers": {
"garmentverse": {
"url": "https://garmentverse-api-807711804803.asia-south1.run.app/.well-known/mcp.json",
"headers": {
"Authorization": "Bearer ${env:GARMENTVERSE_API_KEY}"
}
}
}
}#Safety guarantees
consent_tokenrequired on every request — your agent must explicitly affirm rights to the input person image.- Mode-2 transfers only the garment, never the reference person's face.
- Celebrity face-likeness detection blocks unauthorized identity transfer.
- Minor face detection gates submissions where age estimate < 18.
- Outputs carry C2PA provenance metadata; draft tier is watermarked.
Need a key?
Sign up at the dashboard. 100 credits free, pay-as-you-go after.