iamfaham/investor-data-mcp

Investor Data MCP Server

A Model Context Protocol (MCP) server that exposes tools, resources, and prompts for fetching and analyzing venture capital investor data from a Supabase-hosted OpenVC dataset.

Highlights

• Investor data access: Names, websites, HQ, countries, stages, cheque sizes, thesis
• Flexible search: Filter by investor type, investment stage, country of investment, or HQ location
• Analytics: Stage distribution, database stats, thesis theme analysis, similar investors
• MCP-native: Tools/resources/prompts consumable by any MCP-compatible client

Requirements

• Python 3.12+
• Supabase project

Environment variables (set in your shell or a .env file):

• SUPABASE_URL
• SUPABASE_KEY
• SUPABASE_TABLE

Example .env:

SUPABASE_URL=https://your-project-id.supabase.co
SUPABASE_KEY=your-supabase-anon-or-service-role-key
SUPABASE_TABLE=dec-2024

Setup

Install dependencies (recommended for the included test clients):

pip install -r requirements.txt

Alternatively, install from pyproject.toml:

pip install .

Run the MCP Server Locally

python server.py

On startup, the server prints its MCP endpoint. By default:

• MCP endpoint: http://localhost:8000/mcp

Available Tools

• get_investor_data(limit?: int)
• search_investors_by_criteria(investor_type?: str, stage?: str, country?: str, hq_location?: str, limit?: int)
• get_available_investor_types()
• get_available_countries()
• get_location_search_guide()
• analyze_investment_stages()
• find_investors_by_cheque_size(min_amount?: str, max_amount?: str, limit?: int)
• analyze_investment_thesis()
• get_investor_statistics()
• find_similar_investors(investor_name: str, limit?: int)

Notes:

• country accepts common names/codes (e.g., "USA", "UK").
• Cheque size filtering is string-based and approximate.

Resources and Prompts

• Resource: docs://vc_data_guide
• Prompt: analyze_investor_data(investor_data: str)

Test Locally (LangChain + MCP)

test_client_local.py demonstrates connecting a LangChain agent to the local MCP server via langchain-mcp-adapters and ChatGroq.

Prerequisites:

• Local server running (python server.py)
• .env contains GROQ_API_KEY for ChatGroq

Run:

python test_client_local.py

It exercises multiple tools (investor data, search, analytics) and prints truncated outputs.

Test a Smithery Deployment

test_client_smithery.py validates a Smithery-hosted deployment.

Prerequisites in .env:

• SMITHERY_API_KEY
• SMITHERY_PROFILE
• Optional: GROQ_API_KEY (for tests; otherwise skipped)

Run:

python test_client_smithery.py

It first checks a direct MCP connection to a Smithery URL, then (optionally) runs LangChain agent tests. The Smithery build is configured via smithery.yaml and the container via Dockerfile.

Docker

Build:

docker build -t investor-data-mcp .

Run (provide Supabase env vars):

docker run --rm -p 8000:8000 \
  -e SUPABASE_URL="https://your-project-id.supabase.co" \
  -e SUPABASE_KEY="your-supabase-key" \
  -e TABLE_NAME="table_1" \
  investor-data-mcp

The container exposes the MCP endpoint at http://localhost:8000/mcp.

Data Schema Expectations

Expected table (configurable via SUPABASE_TABLE env var) with columns such as:

• Investor name
• Website
• Global HQ
• Countries of investment
• Stage of investment
• Investment thesis
• Investor type
• First cheque minimum
• First cheque maximum

Update the SUPABASE_TABLE environment variable or selected columns in server.py if your schema differs.

Integration Notes

• Transport is streamable-http; the MCP endpoint is /mcp.
• You can discover tools with an MCP client directly or use LangChain via langchain-mcp-adapters.

Minimal MCP client example (see test_client_local.py for a full example):

from langchain_mcp_adapters.client import MultiServerMCPClient

client = MultiServerMCPClient({
    "investor_data": {
        "url": "http://localhost:8000/mcp",
        "transport": "streamable_http",
    }
})

tools = await client.get_tools()

Troubleshooting

• Ensure SUPABASE_URL and SUPABASE_KEY are set (env or .env).
• Verify the Supabase table (SUPABASE_TABLE env var) exists and column names match.
• Set GROQ_API_KEY to enable LangChain agent testing.

Data Source

The investor data is sourced from OpenVC, a comprehensive database of venture capital investors worldwide. The dataset was last updated in December 2024, providing current information on investor preferences, investment stages, thesis, and contact details.