Skip to content
Dark MatterDark Matter
MCP

MCP Quickstart

Connect an LLM to DMTR via MCP in 5 minutes.

What you will build

You will connect an LLM (Claude, Cursor, or another MCP client) to DMTR in under 5 minutes. By the end, you can ask natural language questions like "What positions do I have?" and receive answers backed by your DMTR registry data. The setup involves installing the MCP server and configuring your client with your DMTR API key.

Prerequisites

  • A DMTR API key (from the DMTR dashboard or sandbox)
  • An MCP-compatible client (Claude Desktop, Cursor, or custom)
  • Node.js 18+ or Python 3.10+ (depending on MCP server runtime)

Install

Install the DMTR MCP server via npm or pip. For Node.js: npm install -g @dmtr/mcp-server. For Python: pip install dmtr-mcp. Alternatively, use the GitHub releases for a standalone binary. The server exposes DMTR as MCP tools and resources.

Configure

Add the DMTR MCP server to your client configuration. For Claude Desktop, edit claude_desktop_config.json and add a server entry with the command and env vars for your API key. For Cursor, add the server to MCP settings. Provide DMTR_API_KEY (or path to credentials) so the server can authenticate to DMTR.

First query

Start a conversation with your LLM and ask: "What positions do I have in the registry?" The LLM will invoke the query-registry tool, the MCP server will call the DMTR API, and you will receive a natural language summary of your positions. Try follow-up questions like "Show me positions over $1M" or "Which positions are transferable?"

Next steps

Edit this page on GitHub

Last reviewed: March 13, 2026

Was this page helpful?

MCP Overview

What DMTR MCP is for and how it connects LLMs to private market data.

MCP Architecture

How the DMTR MCP server exposes resources, tools, and prompts.

On this page

What you will buildPrerequisitesInstallConfigureFirst queryNext steps