Is MCP the solution?

I’ve been using Axure for a looooong time… as in I started with version 5. Now we use 11. My developers are using AI for coding and want to be able to use my prototypes to produce code and/or requirements. So the vision is the UX designer leads with wireframes, uses AI as a co-pilot to build and edit them quickly, and then attaches or captures backend business rules later

“Design-First, Spec-Later” pipeline. They want the specs to be usable by AI for their development work. Basically, the prototype would serve multiple purposes.

  • Designing the interactions for dev for initial collaborations.
  • Gaining buy-in from the client through demonstrating interactivity and various states - they can do this on their own of course through Axure Cloud and adding comments.
  • Post-approval - collaboration with dev and client - working through fringe cases
  • Exporting the specs for AI consumption.

I’m looking to the community and Axure support for ideas on how to get attain these goals. If the right solution is to integrate Figma into the process, that’s fine. I just need some direction and guidance. I have not used Whiteboards - I do know that they can be used through MCP - just not sure it this will achieve the goals.

Thanks!

1 Like

Hi @tallgal ! We are planning to add MCP to RP and that will likely be the best solution as part of this workflow. Currently, the prototype or the Word spec (or screenshots and images copied from RP) could be used by AI as well, but we are viewing MCP as the better solution.

@victor Is there a timeline for that? I noted that MCP is in alpha for whiteboards and documents.

@tallgal We’re actively working on the feature now. I would say within the next few months, hopefully sooner.

@tallgal Hi! MCP for Axure RP is now available in the latest release candidate (11.0.0.4147) which you can download through the updater in RP or from here https://www.axure.com/release-candidate.

Here are the docs for setting up the MCP server and configuring VS Code, Cursor, ChatGPT, Claude, etc. https://docs.axure.com/axure-rp/reference/mcp/

我的AI和我一起试了很多次,能联通,这个http://127.0.0.1:41414/health也显示OK,但是就是无法读到打开的文件,AI说,按文档中的方法,无法看到文件

Hi @carejoe,

My AI and I have tried numerous times; the connection works, and http://127.0.0.1:41414/health shows “OK,” but it simply cannot access the open file. The AI ​​reports that following the methods outlined in the documentation fails to make the file visible.

To start, can you let us know the AI tool you’re using? And can you let us know any specific prompts you’re using with the AI tool? Does your AI at least recognize the Axure MCP as an available tool to use within your current chat?

I just tested the Axure MCP with the Claude desktop application, and Claude was able to recognize my current selection, the interactions the widget has, etc. If possible, please try connecting the Axure MCP to another AI tool and let us know if you run into the same issue. This will tell us if the issue is specific to the AI tool you’re currently using or a possible setting or configuration that’s blocking the Axure MCP from working correctly on your machine.

我用的是openclaw,之后我按你们的推荐尝试一下别的。

Windows / Axure RP 11.0.0.4149: MCP list_documents returns [] with an .rp file open

Hi @Jane_Axure,

I’m experiencing a similar issue to the one reported above. The MCP server connects and its tools are discovered, but an actual list_documents call returns an empty array even though an .rp file is open in Axure RP.

Here are my environment and test results:

  • OS: Windows
  • Axure RP version: 11.0.0.4149
  • Axure RP MCP server: enabled
  • http://127.0.0.1:41414/health returns ok
  • The URL in %LOCALAPPDATA%\Axure\mcp-server.json is http://127.0.0.1:41414/mcp
  • VS Code reports Discovered 10 tools
  • Claude Code reports axure-rp connected with 10 tools
  • The list_documents tool is actually invoked, but the result is []
  • An .rp file is open when the call is made
  • The same result occurs with both Codex and Claude Code

Steps to reproduce in my environment

  1. Open an .rp file in Axure RP 11.0.0.4149 on Windows with the MCP server enabled.
  2. Connect the AI client to http://127.0.0.1:41414/mcp.
  3. Confirm that the Axure MCP tools are discovered.
  4. Invoke list_documents while the .rp file remains open.

Expected result: The currently open .rp document is included in the returned document list.

Actual result: The tool returns an empty array: [].

Could you help determine whether this is a known issue in build 11.0.0.4149 or whether additional configuration is required? Is there a newer build or a workaround that addresses this behavior?

Please let me know which logs or diagnostic information would help investigate why the open document is not being detected.

Thank you.

Hi @jaeyong,

Thank you very much for those details. I was able to reproduce the same issue on Windows (Mac is fine), and our development teams are looking into the issue.

In the meantime, you can install the previous version of the Axure RP 11 beta (4147) as it looks like the issue is specific to build 4149. You can grab that build here.

We’ll update this thread once we have a fix, and please let us know if you run into any other issues.

Hi @jaeyong,

We just released build 11.0.0.4150 which contains a fix for this issue. You can update via Help > Check for Updates from the main menu of Axure RP 11 (with the “Include beta channel builds” selected), or by grabbing the installer directly from: https://www.axure.com/rc.

Please let us know if you run into any issues with the new build. Thanks!

This is fantastic! I’ve only just begun to explore what MCP can do and I’m blown away. I can actually have ChatGPT navigate complex flows and see it navigate my prototype, fill out complex forms, figure out logic, etc. This is a powerful addition! Thank you. I’ll share more as I figure out what deliverables for product owners and developers look like. I understand that Figma offers a lot of functionality for collaboration but I don’t think it’s as powerful as Axure when it comes to building high-fidelity, functional, data-driven prototypes… Has anyone experimented with MCP to generate videos?

Hi @marcbecquey

Would love to hear more about how you’re using the MCP server in your workflow. I haven’t tried generating a video from it but that’s an interesting idea for a deliverable. Are your developers interested in using it directly for coding?