How I Automated Trade Show Tracking With a Second MCP Server (Claude Searches, ContentCal Fills the Calendar)

This is part 3. In part 1 I described how ContentCal grew into the marketing team's operating system, including its trade-show module. In part 2 I connected Claude to the calendar with MCP. This post is about the next use for MCP: finally removing the manual work around trade shows.

The problem: trade shows were entered by hand#

Our marketing team needs to know which trade shows and industry events are coming up, by region. In ContentCal, that lives in the trade-show section, described in part 1.

But someone still had to put the events in. Again and again, someone searched, copied, pasted and formatted. The trade shows were in the system, but they got there by hand.

First idea: scraping, and why it is a trap#

In part 1 I already described the obvious fix: automate event research by scraping. Search, scrape, structure, populate.

Scraping looks great on a whiteboard and is painful in practice:

• Roadblocks. Websites put up bot protection and block automated visitors, and your scraper simply stops getting pages.
• It breaks. A site changes its layout and the scraper returns nothing, or the wrong thing.
• Weak credibility. What comes out is whatever the page said, with no judgement about whether it is a real, relevant event.

A scraper is something you have to keep fixing. I wanted something that behaved like a researcher.

The fix: a second MCP server, just for trade shows#

In part 2 I set up an MCP server so Claude could work with the calendar. For trade shows I set up a second MCP server, built only for this job.

Claude already has its own web search in the Claude app. So the flow became:

1. I ask Claude for the trade shows for a region and period.
2. Claude runs its own web search, as a person would, and works out which events are real and relevant.
3. It builds the results into a sheet.
4. The sheet goes into ContentCal through the connector, straight into the trade-show section.
5. The events fill in by region.

The tools on this server are small and specific: add trade shows, check what is already there, remove trade shows, and read the trade-show context for a workspace. Nothing else.

Claude decides what looks genuine#

The part I like most is the filtering. Claude does not paste in everything it finds. It judges whether each event looks genuine, using things like an official website, dates, a venue and an organiser, and only then loads it.

You can see it in the module above. The Saudi Green Building Forum is loaded with its month and city marked as to be confirmed, and a note that one aggregator lists a single date and it is unconfirmed. An event that cannot be verified still arrives, but it arrives flagged instead of looking certain.

That judgement is not perfect, and I would not pretend it is. Treat the loaded events as a first pass, and spot-check anything important before building a campaign around it. The difference is that the list arrives already filtered, instead of arriving as raw search results.

The second problem: nobody knew who was going where#

There was another gap. Our stakeholders attend different shows, and we usually did not find out until after the fact. Which meant we could not plan their posts in time.

So I built a public page.

• It lists the regions, for example Canada, GCC and Malaysia.
• A stakeholder picks their region and says which show they are attending.
• Each region is connected to its own calendar grid: Canada to Canada, GCC to GCC, Malaysia to Malaysia.

No ContentCal account is needed to use it, which is the point: the people who know where they are going should not have to learn a tool.

What happens when someone submits#

Once a stakeholder says they will attend a show, the rest happens automatically:

1. A card is created on the public grid for that region.
2. The post for the show is added to the calendar, with its image assets.
3. The workspace owners get an email, through the email server I already had, saying who is going to which show.

So the owners learn about it immediately, with the details, and the content is already on the calendar to plan around.

The whole workflow, end to end#

Put together, the trade-show workflow now runs like this:

Nobody retypes a spreadsheet, and the information arrives by itself.

What to be careful about#

• Review what Claude loads. A good filter is not a guarantee. Keep a person at the review step.
• Keep the tool list short. The trade-show server can add, check and remove trade shows, and nothing else. Do not give it more than the job needs.
• A public page is public. Think about what you expose and what a stranger could submit, and decide how you want to handle that.
• Keep the servers separate. One focused server per job is easier to reason about than one server that does everything.

Why this is a good MCP use case#

Trade shows are a great fit for MCP because the work is repetitive, the data is on the open web, and the answer needs judgement, not just copying. A scraper is good at the first two and bad at the third. Claude is the opposite.

If you have a manual research or data-entry job that keeps landing on one person, and you want Claude connected to your own tool in a safe, narrow way, share a short brief through the project form on this site and we can scope it together. You can also read about the AI automation service or browse more case studies.