Directify MCP Tutorial: Run Your Whole Directory From Claude

This morning I needed the URL slugs for every listing added to one of my directories in the last 24 hours, so I could submit them to Google Search Console and Bing Webmaster Tools.

I asked Claude for them. It read the directory, filtered by date, and handed back the two slugs.

I never opened the dashboard. I never exported a CSV. I never sorted a table by created date and squinted at timestamps.

That is the Directify MCP server doing its job, and it is worth understanding properly before you set it up.

Now, saving four minutes on a slug lookup is not interesting on its own. What is interesting is what happens when a four minute task drops to zero effort. You start doing it every day instead of every third week. The tasks you skip because they are tedious are almost always the tasks that compound.

What MCP Actually Is

MCP stands for Model Context Protocol. Ignore the name, it tells you nothing.

Here is the practical version. Without MCP, you ask Claude how to do something and it writes you instructions. You then go and do it yourself. With MCP, Claude has hands. It calls the software directly.

Directify exposes more than 35 of these operations: listings, categories, tags, blog articles, custom pages, organizers, and custom fields. Create, read, update, delete, and bulk create.

So the assistant is not guessing about your site. It reads your live data and writes back to it. That distinction is the entire difference between a chatbot and a workflow.

The Manual Workflow, Honestly Described

I built directories the manual way for two years, so this is not a strawman.

You find a tool worth listing. You open its site. You read the homepage and the pricing page. You write a description. You hunt down a logo URL and a hero image URL. You pick a category. You write an SEO title under 60 characters and a meta description under 155. You open the dashboard, paste eight fields into eight boxes, and save.

Twelve minutes if you are quick and you know the niche.

Fifty listings is ten hours. And hour ten is worse than hour one, because by then you are quietly copying your own phrasing from listing forty. Every description starts to sound the same.

That sameness is the actual cost, not the hours. Google is reasonably good at recognising a directory where 200 listings were filled in by a tired human working from a template. Thin, repetitive listings are why most directories never rank.

Use Case 1: Bulk Listings That Are Not Copy Paste

The obvious win first.

You hand Claude a list of 40 tools with their URLs. It researches each one, writes a description in your directory's voice, assigns a category, drafts the SEO title and meta description, and pushes all 40 in through bulk_create_listings.

The part that matters is not the speed. It is that listing 40 gets the same attention as listing one, because the model does not get bored.

There is a second thing here that saved me from a duplicate content problem. Directify exposes check_listing_exists. So the instruction becomes "check each of these against what is already live, skip anything that exists, and tell me which ones you skipped." You get an actual answer instead of finding out three weeks later that you published the same tool twice under two slugs and split your own rankings.

Use Case 2: The Audit You Never Actually Run

Every directory owner knows they should audit their listings. Almost nobody does, because it means opening 200 records one at a time.

With the MCP connection you ask a question instead:

"Show me every listing with no meta description, or a meta description over 155 characters."

"Which listings have no logo image?"

"Find any listing whose outbound URL is dead."

That last one matters more than people think. Tools shut down constantly. A directory full of links to products that no longer exist bleeds trust with both readers and Google, and you will not notice it happening because nothing breaks visibly on your end.

I ran that check on a Reddit tools directory and found two products that had closed. Rather than delete the listings, I rewrote them to explain what happened and point to a working alternative. Dead link removed, page kept, reader helped. That is a twenty minute job when the audit is free and a job you never start when it is not.

Use Case 3: Editing 200 Listings at Once

Site-wide changes are where the manual workflow really falls apart.

Say you decide every listing needs an affiliate disclosure line. Or you rebrand and the CTA wording changes. Or you realise your meta descriptions all start with the same four words and Google is treating them as boilerplate.

Manually that is a weekend. Through the MCP connection it is one instruction, and the model works through them one at a time.

Same story for category reshuffles. "Move everything tagged scheduling out of Marketing and into Organization Tools, then update the internal links that pointed at the old category." That is the kind of cleanup that sits on a to do list for six months precisely because it is boring and fiddly.

Use Case 4: Site Architecture You Can Actually Reason About

Categories are the skeleton of a directory. Get them wrong and every listing you add makes the problem worse.

The useful move here is asking for analysis before asking for changes. "Read my current categories and my 180 listings, and tell me where the structure does not match the content."

When I did this the answer was uncomfortable and correct. One category had 90 listings and three had four each. That is not a taxonomy, that is a dumping ground with decoration. Splitting the big one into three real subcategories gave me three new pages that could each rank for something specific, instead of one page trying to rank for everything.

Directify supports hierarchical categories through the MCP tools, so the restructure itself was a handful of calls rather than an afternoon of drag and drop.

Use Case 5: The Blog, and Internal Links That Exist

Directory blogs are usually where SEO goes to die. Someone writes ten posts, links to nothing, and wonders why the listings never rank.

Because the assistant can read both your articles and your listings, it can write a post that links to the listings you actually have. Not invented ones. Real slugs, checked against the live directory.

The reverse works too. "Read this new article and add a link to it from every listing it mentions." Internal linking stops being a chore you promise to do later.

This article was written and published through that same connection, which is either a good demonstration or slightly too cute, depending on your mood.

Manual vs MCP, Side by Side

The pattern across that table is not really speed. It is that the boring maintenance work stops being something you avoid.

Where This Breaks

I would rather you hear this from me than find out on your own site.

It will make things up if you let it. Ask for 40 listings with pricing and features and some of that pricing will be wrong, because pricing pages change and models fill gaps confidently. Tell it to pull facts from the live site and to write "not published" when it cannot find something. Then spot check. This is the single biggest risk and it is entirely manageable if you plan for it.

Bulk operations point both ways. The same connection that creates 40 listings in one go can delete them. Be specific about what you are asking for, and be careful with instructions containing the words "all" and "remove" in the same sentence.

Large directories overflow the context window. This one bit me twice while writing this piece. Asking for a full listing dump on a directory with a couple of thousand entries returned about 686,000 characters and blew straight past the limit. The fix is to ask for what you need rather than everything: filter by date, by category, or by page. Worth knowing before you assume something is broken.

It is not strategy. Choosing the niche, deciding the monetisation model, and judging whether a directory is worth building at all is still your job. The MCP connection removes execution friction. It does not tell you what to execute. If you are still working that part out, start with picking a profitable hyper-niche before you automate anything.

Setting It Up

Two paths. The hosted one takes about three minutes.

Step 1. In your Directify account, go to Settings, then API, and generate an API token. Keep it somewhere safe. It is a password.

Step 2, hosted route. Add a custom connector in Claude pointing at https://mcp.directify.app/mcp. It uses OAuth, so you approve access in a browser window and you are done. This is the route I would recommend for almost everyone.

Step 2, local route. If you prefer running it yourself, install with npm install -g directify-mcp and configure it with environment variables. Same tools, more control, marginally more setup.

Step 3. Find your directory ID. It is in the dashboard URL when you are editing your site. You will need it in your first prompt, or you can set it once as a header so you never type it again. Skip this and your first request will come back asking for it.

Step 4. Start with something read only. "List my categories." "How many listings do I have?" Confirm the connection works and that you are pointed at the right site before you let it write anything.

Full configuration details are on the Directify MCP server page.

What I Would Do Next

If you already run a directory on Directify, connect it this week and run one audit. Ask which listings are missing meta descriptions. The answer usually explains a chunk of why traffic is flat, and it costs you one question to find out.

If you have been putting off starting because the data entry looked like a second job, that objection is now considerably weaker than it was a year ago. The research and the writing and the publishing all happen in one conversation.

You can start a directory on Directify here, and the MCP connection is available on directories created during the free trial, so you can test the whole workflow before paying for anything.

I have also written up how I built a directory using Claude Code, which covers the same idea from the developer side if that is more your speed.