Self Checks
1. Is this request related to a challenge you're experiencing? Tell me about your story.
Body:
Hey folks, I've been using Dify heavily lately to build workflows, and I’ve got some thoughts I can’t hold back. I also have a few ideas that could make it "take off"—wondering if anyone else feels the same.
First, the pain points:
DSL is powerful, but the learning curve is no joke. Having to constantly check the documentation breaks my flow.
When working on complex workflows, I often get stuck on "how should this logic be structured?" or "why is this part failing?"—basically figuring everything out manually. It’s exhausting.
Here are a few features I’ve been imagining (wishlist time):
- Two-way DSL ↔ Directory View conversion
Right now, DSL is just a block of code. It’d be awesome if we could one-click convert it into a tree-like directory view to visualize the structure. Even better if dragging and dropping in the directory could automatically sync back to the DSL.
- An official "Dify Expert" Skill (the big one!)
This is what I want the most! Imagine if Dify offered an official Skill (maybe called dify:advisor) that tools like Cursor, Claude, or Codex in Cursor could call directly. This Skill would be a Dify domain expert, and the key is it needs exploration capabilities—something like opsx:exploreor superpower, where it can help clarify requirements and explore possibilities through multi-turn conversations. You describe the gist, and it asks follow-up questions, helps refine the scenario, suggests structures, and even generates DSL drafts. Wouldn’t that be way better than digging through docs?
Oh, and it’d be great if this Skill could also generate an HTML flowchart prototype! When I’m in meetings aligning with teammates, instead of dryly explaining "there’s a condition here, a loop there," I could just share an interactive diagram. Communication efficiency would skyrocket.
- Support real-time workflow editing via MCP for Agents (like Claude)
Sometimes when chatting with an AI, I want to tweak the workflow. If we could use the MCP protocol to just say "remove that approval node," and the AI could make the change live, without me switching interfaces and clicking around—that’d be seamless.
- Support workflow debugging via MCP for Agents
When a workflow breaks, the logs are a headache. If we could pass error stacks and node states to an Agent via MCP and have it analyze "where it’s stuck," "what might be wrong," or even suggest fixes—that’d be a lifesaver.
To sum it up:
The core ask is simple—expose Dify’s capabilities (DSL, workflow editing, debugging) via Skill and MCP, so AI Agents can truly act as our assistants. From actively exploring needs, brainstorming ideas, and drawing prototypes, to live debugging and real-time editing—freeing us from repetitive manual work so we can focus on design and logic.
What do you all think? Anyone else feeling these pain points? Let’s chat!
2. Additional context or comments
No response
3. Can you help us with this feature?
Self Checks
1. Is this request related to a challenge you're experiencing? Tell me about your story.
Body:
Hey folks, I've been using Dify heavily lately to build workflows, and I’ve got some thoughts I can’t hold back. I also have a few ideas that could make it "take off"—wondering if anyone else feels the same.
First, the pain points:
DSL is powerful, but the learning curve is no joke. Having to constantly check the documentation breaks my flow.
When working on complex workflows, I often get stuck on "how should this logic be structured?" or "why is this part failing?"—basically figuring everything out manually. It’s exhausting.
Here are a few features I’ve been imagining (wishlist time):
Right now, DSL is just a block of code. It’d be awesome if we could one-click convert it into a tree-like directory view to visualize the structure. Even better if dragging and dropping in the directory could automatically sync back to the DSL.
This is what I want the most! Imagine if Dify offered an official Skill (maybe called dify:advisor) that tools like Cursor, Claude, or Codex in Cursor could call directly. This Skill would be a Dify domain expert, and the key is it needs exploration capabilities—something like opsx:exploreor superpower, where it can help clarify requirements and explore possibilities through multi-turn conversations. You describe the gist, and it asks follow-up questions, helps refine the scenario, suggests structures, and even generates DSL drafts. Wouldn’t that be way better than digging through docs?
Oh, and it’d be great if this Skill could also generate an HTML flowchart prototype! When I’m in meetings aligning with teammates, instead of dryly explaining "there’s a condition here, a loop there," I could just share an interactive diagram. Communication efficiency would skyrocket.
Sometimes when chatting with an AI, I want to tweak the workflow. If we could use the MCP protocol to just say "remove that approval node," and the AI could make the change live, without me switching interfaces and clicking around—that’d be seamless.
When a workflow breaks, the logs are a headache. If we could pass error stacks and node states to an Agent via MCP and have it analyze "where it’s stuck," "what might be wrong," or even suggest fixes—that’d be a lifesaver.
To sum it up:
The core ask is simple—expose Dify’s capabilities (DSL, workflow editing, debugging) via Skill and MCP, so AI Agents can truly act as our assistants. From actively exploring needs, brainstorming ideas, and drawing prototypes, to live debugging and real-time editing—freeing us from repetitive manual work so we can focus on design and logic.
What do you all think? Anyone else feeling these pain points? Let’s chat!
2. Additional context or comments
No response
3. Can you help us with this feature?