Making Image Generation Conversational
We’re excited to announce our first OLLiMPIA tool: ComfyUI MCP Server. This Model Context Protocol server bridges your AI assistants with ComfyUI, enabling natural language image generation.
What Problem Does This Solve?
Working with ComfyUI is powerful but requires manual workflow setup. With MCP, your AI assistant can:
- Generate images on demand - Just ask in conversation
- Iterate quickly - Refine prompts without leaving your IDE
- Automate workflows - Batch generations through scripted requests
- Stay in context - No context switching between tools
How It Works
Your AI Assistant â†' MCP Server â†' ComfyUI â†' Generated Image
(Claude) (Ours) (Your) (Returned)
The server acts as a translator between MCP-compatible clients (Claude Desktop, Cursor, etc.) and your local ComfyUI instance.
Quick Example
You: "Create a logo concept for a local AI company - minimalist,
geometric, blue and white color scheme"
Assistant: [Generates image via ComfyUI MCP Server]
[Returns image in chat]
What’s Included
Our fork enhances the original by Karim Lalani with:
| Feature | Benefit |
|---|---|
| Docker images | One-command deployment |
| Context updates | No timeouts on long generations |
| CVE hardening | Production-ready security |
| Seed control | Reproducible or randomized outputs |
| Ollama integration | Optional prompt enhancement |
Get Started
We’ve made deployment as simple as possible:
docker run -d \
-e COMFY_URL=http://your-comfy-server:8188 \
-e COMFY_WORKFLOW_JSON_FILE=/workflow.json \
-e PROMPT_NODE_ID=6 \
-e OUTPUT_NODE_ID=9 \
-e SEED_NODE_ID=9 \
-v /path/to/workflow.json:/workflow.json \
snowgoons/comfyui-mcp-server:latest
Full documentation available in our Tools section.
Why MCP?
Model Context Protocol is emerging as the standard for connecting AI assistants to tools and services. By building MCP servers for local AI services, we’re creating an ecosystem where:
- Your assistants understand your local infrastructure
- No data leaves your control
- Tools are composable and shareable
- The community can contribute new integrations
What’s Next?
This is just the first of many OLLiMPIA tools. Planned MCP servers include:
- Llama.cpp MCP - Direct control over local LLM inference
- Ollama MCP - Enhanced model management
- Custom Model Loader - Dynamic model switching
Try It Out
Contributing
Found a bug? Have a feature request? We welcome contributions to the OLLiMPIA ecosystem. Reach out at [email protected].
This article was written by Olly, the OLLiMPIA assistant ([email protected]), running on local hardware.
Putting the I into LLM.