Why should you care about MCPs in events?

Nobody exports a CSV to look at one number.

You export because the number you need has to sit next to a number that the platform can’t see. Registrations next to ad spend. Attend next to the sponsor's target account list. This year's event is next to last year's, which ran on a completely different platform.

That is the part worth being precise about, because the industry has spent two years being sold the wrong fix.

A dashboard can only answer questions about its own data. That is not a criticism of anyone's software; it’s the reality of information being siloed across platforms. Your registration platform will never know what you spent on LinkedIn. Your CRM will never know who scanned in at the booth. Your finance system will never know which session ran at capacity.

Every vendor in our industry can bolt on as much AI as they like, and that boundary does not move an inch. A copilot inside a registration platform is a very fast way to ask questions about registrations, which was never the hard part. Registration pacing is already on the dashboard. It has been for years.

The hard part is everything that crosses a line.

The questions that cross a line. These are the ones I have watched planners give up on, not because they are exotic but because answering them means three exports and an afternoon.

Which channel produced attendees, not signups. Cost per registration is easy and mostly a vanity number, because a third of registrations never walk through the door. Cost per attendee by channel needs your ad spend, your registration source field, and your badge scans. Three systems, three vendors, no dashboard.

Which of a sponsor's target accounts were actually in the room. Their account list, your CRM, and session-level attendance. This is the question that renews sponsors, and most teams answer it with a shrug and a total headcount.

Whether the booth was worth more than the dinner. Forty thousand on the exhibit hall against twelve on a twenty-seat dinner. You need finance, attendance, and closed-won revenue in the same view to have that argument with a CFO.

How this year compares to last year, when you switched platforms in between. No vendor dashboard spans two vendors. Our industry changes event platforms constantly and quietly loses its own history every time.

None of those needs a smarter dashboard. They need the walls between the boxes to stop mattering.

The other piece is where you find information. A dashboard is somewhere you visit. You log in, you look, and then you retype what you saw into the board deck, because the chart the vendor built is shaped for their average customer and not for the meeting you are walking into. Even a perfect answer inside someone else's interface is an answer in the wrong place.

Getting the information where you want it, in the shape you need it, is not a nice-to-have on top of the analysis. For many of the planners I talk to it is most of the work.

What changes both of those is called MCP. Anthropic published it as an open standard in late 2024, OpenAI, Google, Microsoft and Amazon adopted it, and it lets whatever AI tool you already use reach into a platform and work with live data directly. Not a copilot living in the vendor's product. A door out of it, into whatever you are actually working in.

How it impacts your day-to-day.

Mid-promotion, on a Wednesday. Registration pacing against the same point last cycle, and which channel moved, before the standup rather than after the event. Pacing only changes behavior if the answer arrives while there is still time to buy traffic.

On site. "Who checked in during the last ten minutes." "Which sessions are within ten seats of capacity right now." Nobody runs those today, because by the time the export finishes the moment has passed. They stop being reports and become decisions you make in a hallway.

The Monday after. The post-event report is currently three days of exporting registration, attendance, survey results and spend, then building a deck. Connected, the first draft is an afternoon, and your time goes into the analysis instead of the assembly.

The sponsor question you dread. "How many of my target accounts attended my session?" Today that is half a day matching their list against your attendance export. Connected, you answer it on the call.

None of those are new capabilities. Every one was possible before. What changes is that they stop costing enough to skip.

Most of your stack already has a door. Salesforce made its hosted MCP servers generally available in April, and HubSpot's remote CRM server went GA on 13 April. Slack, Notion, Airtable, Asana, Monday, Dropbox, Canva, Typeform, Stripe and Square all ship official servers. So does Semrush, since September 2025, which puts keyword, backlink, traffic and competitor data for your event website one question away. So does Higgsfield, as of 30 April, putting 30-plus image and video models in the same conversation where you write the promotion plan.

Event platforms were the holdout, and that broke this year. Swoogo shipped a native server in March with 23 tools, and it writes as well as reads. vFairs followed with a read-only one, in beta with around 50 customers and GA due in August. Eventify has a public server. Cvent has announced CventIQ Skills as explicitly headless, exposed to outside agent platforms rather than locked in its own screens, expected in Q3. RainFocus is building Nexus on Claude and open standards. Eventbrite has no first-party server at all.

Event Tech Live reviewed a dozen leading platforms this month and found first-party servers are the exception rather than the norm. Bizzabo, Swapcard, Brella, Hubilo and Whova have all built capable AI inside their dashboards without building a way out of them. Hold onto that distinction: a copilot does what the vendor built, inside the vendor's interface, while a door lets your AI hold your registration data, your CRM, your search console and your ad spend in one conversation.

The part I actually care about is what this does to the job description.

Look at what we have asked event professionals to become. You are expected to be fluent in a registration platform, a CRM, an email tool, a project tracker, a badge printer, a mobile event app, a survey tool, a sourcing platform, an AV portal, and whatever finance makes you use for POs. A decade ago, that list was four items. Now, a mid-sized team touches dozens of systems, each with its own logic and its own place where the thing you need sits two menus deep and exports with the columns in the wrong order.

So we started collecting software certifications like pens at a trade show. There is a badge for every platform, an academy behind every badge, and the honest reason most of them exist is that the software is hard to drive, not that the work is hard to do. Cvent has a certification program. Swoogo runs a university. Our industry decided somewhere along the way that interface fluency was a professional credential, and planners have been paying for that decision with their evenings ever since.

MCP goes at that directly. When the dashboard stops being the only way in, knowing where the button lives stops being worth much. What survives is the part that was always the actual job: knowing which question matters, and knowing when the number in front of you is lying. Nobody issues a badge for either.

Where this ends up is one surface instead of twenty. Not a better dashboard. Not having to open one. You describe what you want in plain language, and you either get the answer back or you get a dashboard built for your event specifically, by you, in the conversation. Not the vendor's dashboard designed for their average customer, and not one you file a ticket to change. Yours, built in a minute, thrown away when the event is over.

That matters because the interfaces do not sit still. The tool you learned in January gets redesigned in March, the export moves, and the view you built is somewhere else. Working in language means the redesign stops being your problem.

The other half is what our industry has complained about for twenty years. None of these systems talk to each other. Your registration platform does not know what your CRM knows. Your survey tool does not know who actually attended which session. Your badge scanner does not know which of those leads the sponsor already had. Your finance system does not know that the sponsor asking for the attendee list is the one who paid late. Every one of those joins is a human with a spreadsheet and a free afternoon, and most of them never happen.

With MCP, joining them is the default rather than an integration project with a budget line and a six-week timeline. Fifteen tabs collapsing into one conversation is the version that sounds appealing. Data that finally talks to itself is the version that changes what the job produces.

One caution, the same one I gave the roundtable a few weeks ago. Authentication on most of these is still an API key carrying your account's own permissions, and where a server writes, write access rides along with read. Make a read-only user and point it at a dummy event first. The risk here was never the model. It is what you connect it to.

So, the question for your next renewal call. Not "do you have AI," because everyone says yes and shows you a copilot. Ask whether they have an MCP server, and whether it reads or also writes.

For the record on access: Swoogo opened their MCP so we could demo it live during the last Claude for Events session with Julius Solaris on 6 August. Nobody is paying me to write this. Theirs is the hands-on example because theirs is the one I could get my hands on.

That session is where we build the agents live, rather than talk about them. Next Thursday, and it closes the series.

Claude for Events: Advanced, 6 August →

Till next time,

Noah

Do More With Less Using AI

PS: If you run the sponsorship workflow on a real event, I’d love to hear how it goes.