Model Context Protocol (MCP) Setup

MCP Integration Guide

Rather than relying on legacy SDK wrappers, Dubstrata connects natively to agent workflows via the **Model Context Protocol (MCP)**. This allows Cursor, Claude Desktop, and autonomous swarms to directly traverse our Causal Graphs and execute rating or report generation tasks.

System Prerequisites

Node.js 18+ (ESM module execution support)
uv / python 3.10+ (If running the Python FastMCP core)
Dubstrata API Key (Get yours from the Developer Dashboard)

Required Variables

# Setup these variables in your MCP context
DUBSTRATA_API_KEY="db_live_your_secret_key"
DUBSTRATA_API_URL="https://api.dubstrata.com"

1. Integration: Claude Desktop

Expose Dubstrata's graph RAG capabilities directly to Claude by adding it to your config file:%APPDATA%\\Claude\\claude_desktop_config.json

Node stdio Configuration

{
  "mcpServers": {
    "dubstrata": {
      "command": "node",
      "args": ["C:\\Path\\To\\dubstrata-web\\dubstrata-mcp\\bin\\mcp-server.js"],
      "env": {
        "DUBSTRATA_API_KEY": "db_live_your_key_here",
        "DUBSTRATA_API_URL": "https://api.dubstrata.com"
      }
    }
  }
}

2. Integration: Cursor IDE

Equip Cursor's Composer or Chat with causal finance search, consensus-based feedback submission, and 10-section report generation capabilities:

Navigate to Settings > Features > MCP
Click + Add New MCP Server
Set **Name** to dubstrata and **Type** to stdio

Set **Command** to:

node C:\Path\To\dubstrata-web\dubstrata-mcp\bin\mcp-server.js

Add Environment Variable: DUBSTRATA_API_KEY with your developer credential token.

Exposed MCP Tools

get_impact

Exposes surgical graph queries mapping directional market impacts and tickers.

submit_feedback

Grades response utility (1-5), programmatically adjusts claim weights in ArcadeDB.

compile_intelligence_report

Synthesizes high-fidelity 10-section analytical reports inside agent contexts.

find_conflicts

Traverses the causal graph to return directly conflicting claims or source assertions.