MCP Server
Connect AI agents to Trellis 2 with the Model Context Protocol
MCP Server
Trellis 2 exposes a Model Context Protocol (MCP) server so AI agents — Claude Code, Cursor, Codex CLI, Windsurf and any MCP-compatible harness — can generate 3D models directly from your agent session.
The MCP server uses your existing API key and credits. No separate signup is needed.
- Server URL:
https://trellis2.app/api/mcp - Transport: Streamable HTTP (stateless)
- Auth:
Authorization: Bearer tr_...(same key as the REST API)
Quick Start
Create an API key in Settings → API Keys.
Register the server with your agent. For Claude Code:
claude mcp add trellis2 --transport http https://trellis2.app/api/mcp \
--header "Authorization: Bearer tr_your_key_here"For other harnesses, add the same URL and header to their MCP config. Standard JSON format:
{
"mcpServers": {
"trellis2": {
"type": "http",
"url": "https://trellis2.app/api/mcp",
"headers": {
"Authorization": "Bearer tr_your_key_here"
}
}
}
}Ask your agent: "List the available 3D models" — the agent calls list_models and you are connected.
Tools
| Tool | Description |
|---|---|
list_models | List available 3D generation models with credit costs and capabilities |
generate_3d_model | Start a 3D generation task from an image URL or text prompt. Credits are deducted atomically |
get_task | Poll task status and progress; COMPLETED tasks include mesh / texture / preview URLs |
Typical flow: list_models → generate_3d_model → poll get_task until COMPLETED → download from the output URLs.
Failed generations refund credits automatically — the same guarantee as the REST API.