Connect Your Own MCP Servers to AI Chat and Deep Research Agents
New in AI Analyst Chat and Deep Research Agents: you can connect your own remote MCP servers. Add the endpoint for a data vendor you already use, or for your firm's own internal systems, and the Marvin Labs AI uses those tools alongside the validated primary data already ingested and prepared as part of Automated Data Import.
MCP (Model Context Protocol) is the standard for giving an AI access to external tools and data. A growing number of the providers analysts rely on now ship remote MCP servers. This update lets you bring them into Marvin Labs.
What MCP is, in more depth
MCP is an open standard, introduced by Anthropic in November 2024 and since adopted across the major AI platforms, that defines how an AI application talks to external tools and data sources. Before MCP, every integration between an AI product and a data provider was a bespoke build. With MCP, a provider implements one server and it works with any client that speaks the protocol. The standard is no longer a single vendor's project: in December 2025, Anthropic donated it to the Agentic AI Foundation under the Linux Foundation, co-founded with OpenAI and Block.
An MCP server publishes a set of tools. Each tool has a name, a description of what it does, and a defined input format. For a fundamentals provider, the tools might be "look up a company", "retrieve reported financials for a period", and "search filings". The client, in this case the Marvin Labs' AI copilot, connects to the server, reads that list, and makes the tools available to the model. When your request needs one, the model calls the tool with specific inputs, the server runs it and returns the result, and the answer gets built on top of it.
A remote MCP server is one the provider hosts at a URL, using HTTP as the transport. You authenticate with an API key or bearer token (some providers also support OAuth), and there is nothing to install or run on your side. That is the kind of server you can connect to Marvin Labs.
What you can connect
Any provider that publishes a remote MCP server can be connected, whether it is a market data vendor, a fundamentals provider, or your own internal systems. The examples below are the ones analysts ask for most often, but the list is not fixed: if a provider ships a remote MCP endpoint, you can add it.
Typical applications for analysts are:
-
Market data: FactSet ships an official remote MCP server covering fundamentals, consensus estimates, ownership, pricing, and events. S&P Global 2025 exposes Capital IQ Pro data through an MCP server built on its Kensho LLM-ready API. Both are entitlement-gated, so you connect them with your existing subscription.
-
Fundamentals: Daloopa runs a remote MCP server that returns source-linked fundamentals across thousands of public companies, every datapoint hyperlinked back to the filing it came from. Fiscal.ai exposes its financial data terminal the same way, with OAuth or API-key auth. For lighter, API-first coverage, Financial Datasets and Alpha Vantage both offer hosted MCP servers covering prices, statements, and filings. Any MCP you add here comes in addition to the fundamentals Marvin Labs provides outright.
-
Your own internal systems: If your firm runs its own data warehouse, note database, or proprietary model behind an MCP endpoint, you connect it the same way you connect a vendor.
How Marvin Labs uses connected servers in AI Analyst Chat and Deep Research Agents
Once a server is connected, AI Analyst Chat and Deep Research Agents handle the rest automatically. If a request needs data from a connected MCP, the AI calls the relevant data provider and integrates its output into its context.
Validated sources come first
Marvin Labs has a clear preference order. When a request can be answered from the filings, earnings calls, and press releases it ingests through Automated Data Import, it answers from those. That content is validated, linked back to the source passage, and consistent across every feature in the platform. Connected MCP servers do not override it, and connecting one does not change how primary financial content is sourced.
MCP servers exist to bring in data the platform does not already have: a vendor dataset you subscribe to, your firm's internal numbers, a proprietary model. The copilot reaches for a connected server when your request needs that additional data, and it attributes the output to the server it came from.
A connected server is an additional tool, not a replacement for the primary document pipeline. When the copilot uses one, the answer shows the tool call and its output so you can tell what came from your connected data and what came from primary documents. You decide which vendors and internal systems to trust.
Adding a server
You add a server in settings with two pieces of information: the server URL and an auth token from the provider. Once saved, the server's tools are available across your chats and agents until you remove it.
Connecting your own MCP servers is available on all paid plans and to evaluation users.
Quick Start
Try it now
- Get a remote MCP server URL and auth token from a provider you use (or your own internal endpoint)
- Add it in Marvin Labs settings
- Make a request in AI Analyst Chat that the connected server can answer, and watch the tool call appear inline
Frequently asked questions
Alex is the co-founder and CEO of Marvin Labs. Prior to that, he spent five years in credit structuring and investments at Credit Suisse. He also spent six years as co-founder and CTO at TNX Logistics, which exited via a trade sale. In addition, Alex spent three years in special-situation investments at SIG-i Capital.
