Summary

Environment variables are enough — no config file needed on a cloud platform.

Being able to pass the base URL and the secret keys directly as environment variables is what makes this deployable somewhere you cannot drop a YAML file. Otherwise the model is simple and holds up: one App Secret Key per workflow, listed in order, and whatever you built in Dify becomes callable from the assistant you already use.

What it is

A bridge that turns Dify workflows into MCP tools. Each Dify App Secret Key you list becomes a workflow the client can invoke.

What you get

Your Dify workflows callable as tools from any MCP-capable client
Several workflows at once — the SK list is comma-separated, and each key typically maps to a different workflow
Two ways to configure: environment variables, or a config.yaml with dify_base_url and dify_app_sks
Cloud-hosted or self-hosted Dify, since the base URL is yours to set

Requirements

Uv or uvx. Set DIFY_BASE_URL to your Dify API base — https://cloud.dify.ai/v1 for the hosted service — and DIFY_APP_SKS to a comma-separated list of App Secret Keys. Alternatively write a config.yaml and point CONFIG_PATH at it. Run with uvx --from the repository git URL and dify_mcp_server, or clone and run it with uv.