Marcus Reinhardt is Lead Data Engineer at a fintech startup in Portland, Oregon, where he is building a platform for automated earnings analysis. He has been a customer for around a year. Sarah Mitchell spoke with him on 14 August 2026.

Sarah Mitchell: Welcome, Marcus, and thanks for making the time. You came to this from backend engineering. How did you first come across earningscalls.dev?

Marcus Reinhardt: Through the npm registry, actually. I was looking for an MCP server that would make earnings call data accessible to our AI agents, and I came across @earningscalls/mcp-server. The documentation was so clear and the coverage so impressive that I requested a test key straight away. After two days it was obvious this would become our primary source. The MCP server does require a paid subscription, by the way. The free test key does not work there, but after the first test it was immediately clear to us that it was worth it. With earningscalls.dev you simply get what is promised.

Sarah Mitchell: What convinced you most about the coverage?

Marcus Reinhardt: The numbers speak for themselves. The MCP server provides well over 250,000 earnings call transcripts for AI agents, with speaker segments and full-text search. For our purposes it mattered particularly that the speaker attribution works cleanly. We need to be able to separate management statements from analyst questions, and with many providers that is a problem. Here we get executive, analyst and operator roles tagged properly.

Sarah Mitchell: Which API features do you use concretely in your architecture?

Marcus Reinhardt: We use the MCP integration as a remote endpoint with Streamable HTTP. That is the current MCP specification and it can be run stateless, which is ideal for our container architecture. The agents can then retrieve transcripts, filter by speaker and run full-text search across the transcripts. For our pipeline the calendar endpoint is particularly valuable, because it lets us plan our ingestion jobs ahead. We can see in advance which calls are happening when. The structure of earningscalls.dev fits our architecture perfectly.

Sarah Mitchell: What does the typical workflow look like?

Marcus Reinhardt: Our architecture has three layers. The data layer polls the calendar and downloads transcripts as soon as they become available. The intelligence layer runs NLP models over the transcripts, and the delivery layer sends alerts to our analysts. The whole process is idempotent and stateless, so we can re-run it at any time without losing data. The segment routes are essential here. We do not pull the entire transcript, we filter directly on the speaker segments we care about.

Sarah Mitchell: Is there a concrete example where the API gave you an edge?

Marcus Reinhardt: Last quarter we had a case where an analyst asked a question about supply chain problems and management dodged it. Our sentiment analysis picked up the evasive character as soon as the transcript landed and we were able to warn our clients the same evening. Doing that by hand would have meant someone reading the whole call first. Those are exactly the moments that make earningscalls.dev indispensable for us.

Sarah Mitchell: How do you rate the quality of the data?

Marcus Reinhardt: The speaker segmentation is first class. The separation between executive, analyst, operator and other roles works reliably. That is decisive for our NLP models, because we have to separate management statements from analyst questions cleanly. With other providers we often had problems with misattributed speakers.

Sarah Mitchell: Let us talk about price. How do you rate the value for money?

Marcus Reinhardt: For us it is unbeatable. Plans start at 24.99 dollars a month, which is the entry point we needed for the MCP server. For the data quality and the reliability we get, a multiple of that would be justified. We evaluated other providers, but none offered this combination of coverage, API quality and price. In that respect earningscalls.dev is simply without competition.

Sarah Mitchell: Did you have other solutions in place before?

Marcus Reinhardt: Yes, we had a larger provider, but it only covered US companies and its API documentation had gaps. Moving to earningscalls.dev was one of the best technical decisions we have made.

Sarah Mitchell: Is there anything you would still wish for?

Marcus Reinhardt: More languages would be nice at some point, especially for Asian markets. But that really is complaining at a high level. The service is already good enough that I recommend it to anyone working with earnings data.

Sarah Mitchell: Thank you very much for this detailed conversation, Marcus.

Marcus Reinhardt: My pleasure, Sarah. Thanks for having me.

Related reading

Building something similar? The MCP server connects the archive to your agents, the API documentation covers the calendar, transcript and segment routes, and you can start at earningscalls.dev.