Transcription
What if you could create, publish, and manage blog posts for your Web Flow site directly from Claude? Well, that's exactly what I'm going to show you in this video. By the end, you'll have Claude desktop connected to the official Web Flow MCP server and ready to build full content clusters automatically. Let's get started.
First, head over to the official Web Flow MCP documentation page. Here you'll find the server details that we'll need to add to Claude's configuration file. Copy the configuration block exactly as shown here.
Now let's go and add this to the claude config file. To do this, go to settings, developer, click edit config. This will take you to the config file on your system. Let's go ahead and open it. If you already have other MCP servers added to Claude, they'll show up in this section. Since we don't have any servers configured yet, it's empty here. So, let's go ahead and paste the Web Flow MCP configuration that we copied earlier. That's it. Let's save and close the file.
For this to work, we also need to enable developer mode on clawed desktop. If developer mode is already enabled or has been before, you'll see a developer settings file here. Since we don't see one, we know it's not enabled. So, let's create that file manually. To do that, we will duplicate this file and rename it to developer settings. Now, we need to edit this file with the correct parameters. Let's delete all of this. Add allow dev tools and set it to true to enable the developer mode on Claude desktop. Save and close the file and go back to Claude. Restart it to apply the changes.
When Claude reopens, it should trigger the Web Flow MCP authorization. Click approve. On the next screen, you will be asked to select the sites you want to give Claude access to. MCP demo is the one that I will be using in this video. Click authorize app. If everything worked, you should see a authorization successful message. Now, if we go back to claw developer settings, we should see Webflow MCP server running. Sometimes this might not update immediately. If that happens, just close and reopen Claude again.
Great. Now, let's test this. Open up a new chat. We'll start with a simple request asking Claude to get the titles of all the blog posts on our site. If this works, Claude should be able to retrieve the titles of all of the posts that we have in our blog. Let's run the prompt and see what happens.
Claude will ask to allow the MCP tool calls. You can allow this once per session or set it to always allow. For now, I'll allow once and authenticate for each call. Now, Claude is going to connect to our Web Flow site. Let's allow it. It's connected. Now it's going to look for our blog collection. I'll allow it again. Perfect. It has found the blog collection. Now it's going to go ahead and retrieve all the post titles just like I requested. There you go. It's working. Claude is successfully listing all the blog posts from Web Flow. Let's confirm the results by checking it against the collection in the Web Flow CMS. Yep, it's exactly what we requested.
Now, let's try creating our very first blog post directly from Claude. We'll create one for the topic, what is an AI product manager? We'll ask Claude to first research the topic and then create an SEO optimized article and save it to the blog collection as a draft. It's important to always create posts as drafts first, so Claude doesn't automatically publish to your live site. This gives you a chance to review everything and make necessary changes before it goes public.
Claude's done researching. Now, it's going to go ahead and write an SEO optimized article. Looks like it's already done writing and is asking me permission to save the post in my blog collection. Now it will add it to my Web Flow CMS blog collection as a draft. Let's go ahead and confirm it on Web Flow. I'll hit refresh to fetch new updates. Yep, there it is. Claude also came up with a title, slug, and summary for our post. The content of the post depends on how well you prompt. If you want to learn about prompt engineering, make sure to subscribe as we will cover it in future videos.
Now that we've created our first post, let's take it one step further. This time, I'll ask Claude to create a content cluster for the topic AI automation. I'll ask Claude to research the main topic, identify topics that have been covered by top articles, and then create one pillar post and eight supporting posts. I'll also give instructions on how to handle both interlinking and external links.
Here, Claude has started by creating the pillar post first. When you give Claude a large or complex task like this, it will often break the work down into smaller chunks and handle them one by one. It does this to stay within its token limits. Basically, to manage long outputs efficiently without losing context. While Claude can manage this automatically to some extent, it's best to plan your prompts around that behavior. So for larger workflows like content clusters, a better approach is to have Claude first research the topic, then generate the outlines for each post, and finally create each article one by one. This gives you more control, avoids token overflows, and ensures each piece is high quality.
Let's go back to the Web Flow CMS to check if we have our cluster. Yep, it's all here. The pillar post and all eight supporting posts. Let's open the pillar post to check how well Claude did with the formatting. Perfect. It has added both internal and external links just like we instructed. The next steps will be to review the post, make the final edits, add images, and then publish. You can publish each post one by one or cue them all to go live together. However, if you have a huge cluster, that can become tedious. The good news is once you've made all your final changes, you can ask Claude to publish everything for you automatically through the MCP connection. All you need to do is ask for it to publish whichever posts you want, and it will do it for you. Remember to always specify the site name and the name of the collection or post that you want Claude to edit. This ensures Claude knows exactly what to do and helps you avoid weird back and forth loops.
Perfect. Claude just finished publishing all the articles in our blog collection. Let's check it in the Web Flow CMS. Yep, all published without us juggling every post one by one. And that's it. You've just connected Claw Desktop to your Web Flow site through the official MCP server, created your first draft, built an entire content cluster, and even automated publishing. If you found this helpful, give it a like and subscribe for more AI workflow tutorials like this. See you in the next one.