# model-compose MCP server

> Build and deploy declarative AI pipelines, autonomous agents, and MCP servers using a single YAML configuration file.

- Repository: https://github.com/hanyeol/model-compose
- Page: https://hostedmcp.org/servers/hanyeol/model-compose
- Transport: —
- Stars: 119
- License: MIT

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `OPENAI_API_KEY` | no | no |  |

## FAQ

### Can I run local models with model-compose?

Yes, you can run open-weight models on your own hardware using PyTorch, vLLM, or llama.cpp.

### How do I turn a workflow into an MCP server?

You can convert any workflow into an MCP server by changing the controller adapter type to mcp-server in your YAML file.

### Does it provide a user interface for my workflows?

Yes, every workflow automatically receives a browser-based UI using Gradio, which handles inputs, outputs, and streaming.
