From bottleneck to self-serve: 6 Mixpanel MCP customer stories

Ask any team where their data lives and they'll point you to the right tool. Ask them how fast they can get an answer out of it, and the story gets more complicated.

There's usually a person in the middle. A query that needs building. A few hours, or a few days, between the question and the answer.

That's not a data problem. It's an access problem. And it's the one Mixpanel MCP was built to close.

Since we opened the public beta, teams across the world have been connecting MCP to the AI tools they already use and building workflows we didn't anticipate. What follows are six of their stories. None of them are using it the same way. All of them ran into the same bottleneck before implementing it.

Niek Schreurs is the growth lead at Ditto, an 11-person healthcare startup in Europe growing at around 5% week-over-week. The data was all in Mixpanel. The problem was that Niek was the only team member using it.

Every question from the team, whether about activation, post-ship impact, or conversion rates, came to him. He'd context-switch into Mixpanel, run the analysis, cross-reference the Jira ticket, compare before-and-after windows, and write it up in Slack. The entire workflow took 45 minutes per release. Impact reviews were often delayed or skipped entirely when he was stretched.

So he built two Slackbots on top of MCP. Porygon handles analytics questions in natural language. Anyone on the team asks a growth question and gets a data-backed answer in seconds, including automatic causal checks to confirm whether a downstream metric actually followed or just reflected a channel mix shift.

Gengar handles feature launches. Seven days after a feature ships, it fires a Slack reminder, pulls the Jira ticket for context, runs the before-and-after comparison in Mixpanel, and posts the verdict in a thread. Now that it's setup, it just runs without anybody asking for it.

"Analytics went from a platform most of the team ignored to a shared capability everyone uses in Slack. People who never opened Mixpanel can now get insights by chatting to a bot."

That's the spine of every story in this piece. The data was always there. The path to it was the problem.

DANA Indonesia is one of the largest digital payment platforms in Southeast Asia. They have serious engineering resources and yet they ran into the exact same bottleneck.

Mixpanel queries were owned by domain experts. If you needed data, you found the right person and waited. Ricardo Suranta, Head of Front-End Engineering, described it plainly: one person's bandwidth became everyone's blocker.

So his team built on top of MCP to cut that dependency entirely. Here's how Ricardo described what that looks like in practice:

"We have two agents built on top of Mixpanel's MCP. Our Mixpanel Agent is the go-to for anyone who wants to explore data in natural language. It combines MCP and Mixpanel's API to investigate user-reported issues from multiple angles, giving a much richer picture than either could alone."

The second agent, the Error Master Agent, focuses specifically on triage. It queries error trends through MCP, cross-references their Notion knowledge base, and surfaces error definitions, root causes, and ownership in a single message.

The result: "Now anyone on the team can query Mixpanel data in natural language, directly from chat. No waiting, no bottlenecks, no dependency on the domain owner."

The same shift occurred at the VP level. Randi Waranugraha, VP of Engineering at DANA, ran into a different version of the same problem. His team was manually opening dashboards, exporting data into spreadsheets, and rebuilding reports every time leadership needed an update.

When App Store ratings dropped unexpectedly, investigating the spike required a lot more time and effort from a specialized engineer. With MCP, agents now analyze the trend, reconstruct user journeys, cross-reference internal documentation, and surface potential root causes automatically.

"Instead of spending hours digging through dashboards to understand an incident, today we can ask a question and get the full story, from user impact to potential root cause, in minutes."

Two people, same company, same bottleneck dissolved in two different directions. When you remove the access constraint, the whole organization moves differently.

Zohar Amouyal at SKIO Music took this a step further. His team built eight AI agents, each with direct access to Mixpanel through MCP. They query autonomously, around the clock, with no human prompting them.

Recently, one of those agents flagged that monthly churn had hit 16%, three times the SaaS benchmark, before anyone on the human team noticed. It pulled the cohort data, cross-referenced support ticket volume, and surfaced that billing confusion was driving more than 40% of cancellations. That insight came from a routine check running at 5:30am.

"The biggest unlock is that our AI agents now have the same data literacy as a senior analyst."

Follow that far enough and removing the bottleneck stops being about faster answers to the questions you already ask. It starts surfacing the ones nobody thought to ask.

​​Said Hadjiat at Graffiti didn't build MCP into his team's internal workflow because he opted to build it into his product instead.

Graffiti is a B2B SaaS platform serving brand managers at large consumer companies. Those users aren't data analysts. They don't want to log into Mixpanel, learn the UI, or manage separate credentials. They want answers.

With MCP, Graffiti piped natural language questions from their dashboard chatbot through a single protocol layer, giving users access to Mixpanel's full query engine, including segmentation, funnels, and cohorts, without any of the underlying complexity.

Here's what that looks like for one of their brand managers on a typical day:

"When a brand manager wanted to understand why scan rates dropped in a specific retail chain, the process was brutal: message the data team, wait for someone to build the query in Mixpanel, get a screenshot or export back, ask a follow-up, and repeat. A simple investigation could take a full day across two teams. Now, that same brand manager types 'Compare scan-to-chat conversion for Carrefour vs Intermarché over the last 3 months' directly in our dashboard chatbot, and gets a formatted answer in seconds."

For teams evaluating whether to build a similar integration, Said is equally direct about what MCP replaced on the engineering side:

"Before MCP, we had two options: either redirect users to Mixpanel's dashboard, which required training and separate credentials, or build custom integrations against Mixpanel's REST API, manually handling segmentation queries, funnel logic, cohort filtering, and all the query-building complexity ourselves. With the MCP server, we pipe natural language questions through a single endpoint that gives us access to the full Mixpanel query engine without reimplementing any of it."

The proof came from their client Bioderma. After seeing the chatbot in action, Bioderma organized an internal presentation for their top management to showcase it, not as a future roadmap item, but as something already running in production. Hadjiat continues:

“When your customer turns your feature into their own internal evangelism moment, you've built something that matters beyond your own use case.”

Maria Sirotkina at Spritz Finance uses MCP for something quieter than autonomous agents or embedded chatbots. She uses it to build better product roadmaps.

Before, her team looked at patterns in Mixpanel and formed hypotheses. The data was there, but pulling enough of it to feel confident in a prioritization call took time that rarely lined up with the pace of decision-making. Roadmap calls leaned on instinct more than evidence.

With MCP, that changed:

"When prioritizing roadmap items, we used to look at the patterns and form hypotheses. Now, we're able to bring new data points and make the story even more complete, and have more confidence."

She described a recent example: conversions surged and no one could explain why but MCP surfaced the answer. Users from one specific country had suddenly discovered the product. Without that visibility, the team would have kept searching for an explanation they already had access to.

Better questions lead to better roadmaps and MCP makes those key questions cheaper to ask.

Greg Stoutenberg at Brainforge isn't a Mixpanel end-user in the traditional sense. As a data and analytics consultant, he sees the same bottleneck across many companies, not just one.

During a recent client engagement, he connected MCP with Claude, pulled Mixpanel data alongside external experimentation documents, and generated a 14-page strategic roadmap for an ecommerce activation funnel in minutes. Work that would have taken the better part of a day.

The time savings are real, but Greg's sharpest observation cuts deeper than efficiency:

"I don't think anyone has ever actually wanted to look at a dashboard. What they really want to do is improve revenue."

Every team and story was after exactly that. The 11-person healthcare startup, the fintech with millions of users, the music tech company running autonomous agents, the B2B SaaS embedding analytics into their own product, the PM building tighter roadmaps. None of them set out to get better at using a dashboard. They set out to move faster, decide smarter, and ship better products. 

MCP just closed the distance between the data they already had and the answers they wanted.

Mixpanel MCP is available for all customers in every region and existing customers can turn it on in Org Settings. If you’re new to Mixpanel you can get started free and connect MCP to the AI tools you're already using like Claude, ChatGPT, Gemini, Cursor, Notion, and more.

Your data already has the answers. The only remaining question is, how fast you can get to them?