You want your AI assistant to answer a question about an earnings call. So you go find the transcript, open it, scroll to the right section, copy a wall of text, paste it into the chat, and then ask your question. If the answer spans two calls, you do it twice. If you need a different company tomorrow, you start over.

That workflow is slow, error-prone, and caps out fast. You can only paste so much before you hit a context limit, and you have no way to search across thousands of calls. The AI is smart, but it's blind to the data unless you hand-feed it.

An earnings call MCP server fixes this. Instead of pasting transcripts, you connect your assistant to a live data source once, and from then on it fetches exactly what it needs, when it needs it. This guide explains what that means in plain English, how it works, and how to set one up.

MCP in 30 seconds

MCP stands for Model Context Protocol. It's an open standard that lets AI assistants like Claude connect to external tools and data sources in a consistent way.

Think of it like a USB port for AI. Before USB, every device needed its own special cable. MCP is the universal connector: any MCP-capable client can plug into any MCP server and immediately know what tools are available and how to call them.

An MCP server exposes a set of tools (search this, fetch that, list these). An MCP client is the app you're already using: Claude Desktop, claude.ai, Claude Code, or Cursor. When the two connect, your assistant can call those tools on its own, mid-conversation, without you copying anything.

So what is an earnings call MCP server, specifically?

An earnings call MCP server is an MCP server whose tools are all about earnings call data. It gives your AI assistant direct, on-demand access to earnings call transcripts and the structured information around them.

Instead of you being the middleman between the transcript and the model, the model queries the data itself. Ask a question in plain language, and the assistant decides which tool to call, retrieves the relevant transcript or speaker segment, and answers with the actual text in front of it.

This is what people mean by MCP for earnings data: a standardized bridge between your AI and a real transcript database, so the AI can look things up instead of guessing or relying on stale training data.

If you want the broader background on the protocol and how it applies here, the MCP explainer page walks through it in more detail.

How the earningscalls.dev MCP server works

earningscalls.dev runs a hosted, remote MCP server. You don't install anything or run a server yourself. You connect to a URL, and the transcripts, search, and metadata are all served from our infrastructure.

Under the hood it's backed by a large, current dataset:

Those speaker segments matter more than they sound. Every segment is tagged by role, so the server knows who was talking: executive, analyst, operator, attendee, or shareholder. That's what makes a question like "what did the CFO say" actually work. The assistant can filter to executive segments instead of scanning the whole call.

The tools it exposes

The server exposes roughly 13 tools. You never call these by hand; your assistant picks the right one based on your question. In plain terms, they let you:

Because the tools are structured, the assistant can chain them. It might look up a company, find its latest call, then pull the executive segments, all from one question you asked.

What you can actually ask

Once connected, you talk to your assistant normally. Here's the kind of question that just works:

"What did NVIDIA's CFO say about data center growth in their last call?"

Behind the scenes, the assistant resolves the ticker, finds the latest call, fetches the executive speaker segments, and answers with quotes from the real transcript. You didn't paste anything.

Other things people ask:

The point is that the AI does the fetching. You do the thinking.

Who it's for

An earnings call MCP server is useful if you spend any real time reading or reasoning over calls:

If you already live in Claude or Cursor, connecting an earnings call MCP server means your existing assistant suddenly understands earnings calls without any custom code.

How to get started

The recommended path is the hosted remote connector. It's the fastest, and the access token is embedded right in the URL, so there's no separate API key header to manage.

Step 1: Generate your connector URL

Log in and go to Dashboard → Connectors → Generate. You'll get a personal connector URL that looks like this:

https://earningscalls.dev/u/mct_xxxx/mcp

Keep it private. That URL is your access; treat it like a password. You can manage it any time from your dashboard.

Step 2: Add it to your client

Claude Desktop or claude.ai: open Settings → Connectors → Add custom connector, then paste your URL.

Claude Code: run one command:

claude mcp add --transport http earningscalls "https://earningscalls.dev/u/mct_.../mcp"

Cursor and other MCP-capable clients follow the same pattern: add a custom HTTP MCP connector and paste the URL.

Step 3: Ask a question

That's it. Ask your assistant something about an earnings call and watch it call the tools on its own. No copy-paste, no context juggling.

If you'd rather build against the raw API, the API docs cover every endpoint. And if you want a step-by-step with screenshots, see How to connect earnings call transcripts to Claude with MCP.

MCP vs. the API: which do you need?

Short version: the MCP server is for AI assistants; the REST API is for code you write yourself.

If your goal is to chat with an assistant that can read calls, use MCP. If you're building a pipeline, dashboard, or backend service, the REST API gives you the same data with full programmatic control. Many people use both. For a full breakdown, read MCP vs. REST API for earnings call data.

What it costs

Reading transcripts on the website is free. Anyone can browse and read calls without paying.

API and MCP access is where the paid plans come in, because that's live, programmatic data access:

Each tool call your assistant makes counts as a request, so a single question that chains a few lookups uses a few requests. For most individual users, Pro is plenty. See the full details on the pricing page.

The bottom line

An earnings call MCP server turns your AI assistant from something you feed transcripts into something that goes and gets them itself. You connect once, then ask questions in plain language and get answers grounded in the actual words executives and analysts said, across 246,000+ transcripts and 12,000+ companies.

No more copy-pasting. No more context limits. Just ask.

Ready to try it? Generate your connector URL in the dashboard or check the plans on the pricing page and connect your assistant in a couple of minutes.