RunComfy MCP
MCP server for RunComfy's Serverless API (ComfyUI) that enables managing deployments, running inference, and retrieving results from AI assistants like Claude and Cursor.
README
RunComfy MCP
MCP server for the RunComfy Serverless API (ComfyUI). Manage deployments, run inference, and retrieve results from AI assistants like Claude, Cursor, and Windsurf.
Endpoint: https://mcp.runcomfy.com/mcp
Docs: docs.runcomfy.com/mcp
What it does
10 tools that mirror docs.runcomfy.com/serverless 1:1:
| Category | Tools |
|---|---|
| Deployment management | list_deployments, get_deployment, create_deployment, update_deployment, delete_deployment |
| Inference | submit_request, get_request_status, get_request_result, cancel_request |
| Advanced | call_instance_proxy |
Quick setup
Claude Code
claude mcp add runcomfy \
--transport streamable-http \
https://mcp.runcomfy.com/mcp \
--header "Authorization: Bearer <YOUR_RUNCOMFY_TOKEN>"
Cursor
Add to .cursor/mcp.json:
{
"mcpServers": {
"runcomfy": {
"url": "https://mcp.runcomfy.com/mcp",
"headers": {
"Authorization": "Bearer <YOUR_RUNCOMFY_TOKEN>"
}
}
}
}
Windsurf
Add to Windsurf Settings > MCP:
{
"mcpServers": {
"runcomfy": {
"serverUrl": "https://mcp.runcomfy.com/mcp",
"headers": {
"Authorization": "Bearer <YOUR_RUNCOMFY_TOKEN>"
}
}
}
}
Get your API token from your Profile page.
Architecture
MCP Client ──Bearer token──> Cloudflare Worker (/mcp)
│
▼
Cloudflare Container
(Python FastMCP app)
│
▼
api.runcomfy.net
(using caller's token)
- Cloudflare Worker (
src/index.ts) — thin proxy: CORS, body size check, forwards the caller's token to the container. No auth logic —api.runcomfy.nethandles authentication. - Python container (
server.py) — FastMCP app with 10 tools. Uses the caller's token (forwarded viax-runcomfy-user-tokenheader) for all outbound API calls. Each user sees only their own deployments. - Cloudflare Container auto-starts on first request, sleeps after 10 minutes idle.
Project layout
src/index.ts Cloudflare Worker entrypoint
server.py MCP tool definitions (10 tools)
runcomfy_client.py RunComfy API client (serverless endpoints)
container_app.py ASGI middleware (request IDs, token forwarding)
container_entrypoint.py Uvicorn startup
container_runtime.py Env validation, structured logging
wrangler.jsonc Cloudflare Worker + Container config
Dockerfile Container image
.env.example Local dev config
Local development
# Python 3.11+
python3.11 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env
# Set RUNCOMFY_API_KEY in .env
python -m container_entrypoint
Local endpoints:
http://127.0.0.1:8000/healthzhttp://127.0.0.1:8000/mcp
In local mode (no Worker), the Python app uses RUNCOMFY_API_KEY from .env for all outbound calls.
Deploy
Requires Cloudflare Workers Paid plan with Containers enabled.
npm install
# Set the API key secret (one-time)
npx wrangler secret put RUNCOMFY_API_KEY
# Deploy
CLOUDFLARE_ACCOUNT_ID=<your-account-id> npx wrangler deploy
The MCP endpoint goes live at https://mcp.runcomfy.com/mcp (custom domain configured in wrangler.jsonc).
Environment variables
Worker secrets (set via wrangler secret put)
| Name | Required | Description |
|---|---|---|
RUNCOMFY_API_KEY |
Yes | Fallback API key for the container |
Worker vars (in wrangler.jsonc)
| Name | Default | Description |
|---|---|---|
CONTAINER_INSTANCE_NAME |
runcomfy-unified |
Durable Object instance name |
CONTAINER_STARTUP_TIMEOUT_MS |
15000 |
Max wait for container start |
CONTAINER_PORT_READY_TIMEOUT_MS |
30000 |
Max wait for port ready |
MCP_MAX_BODY_BYTES |
1048576 |
Max request body size |
RUNCOMFY_SERVERLESS_BASE_URL |
https://api.runcomfy.net |
Serverless API base URL |
Local dev (.env file)
| Name | Required | Description |
|---|---|---|
RUNCOMFY_API_KEY |
Yes | Your RunComfy API token |
RUNCOMFY_SERVERLESS_BASE_URL |
No | Override base URL (default: https://api.runcomfy.net) |
RUNCOMFY_MCP_MOUNT_PREFIX |
No | Path prefix for MCP mount (default: empty) |
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。