Connecticut is one of the few states whose business registry includes a NAICS industry code for new companies. From Oct 1 to Oct 23, 2026, the Secretary of the State's open data on data.ct.gov recorded 3,862 new business registrations, and 346 of them were in construction (NAICS 23). The top codes were residential remodelers (236118, 125 new companies), other specialty trades (238990, 29), finish carpentry (238350, 25), commercial building construction (236220, 24), electricians (238210, 23) and painters (238320, 19). Bridgeport, Waterbury, Stamford, Danbury and New Haven led by city.
Because the industry code is right there, this is a great dataset to hand to an AI agent. Instead of writing SoQL, you ask in plain English: "new electrical contractors in Fairfield County this month." Here's how to set that up with Apify's hosted MCP server.
Step 1: Connect the tool
Add this to your MCP client's config (any client that supports remote MCP servers over Streamable HTTP):
{
"mcpServers": {
"apify": {
"type": "http",
"url": "https://mcp.apify.com/?tools=fetch-actor-details,loomwise/connecticut-new-business-registrations"
}
}
}
On first connect, the client opens a browser window to sign in to Apify (OAuth), so no token sits in the file. The tools parameter loads just this one Actor plus fetch-actor-details, which lets the agent read the input schema and pricing before it runs anything.
Step 2: Ask
Using the Connecticut new business registrations tool, find companies registered in the last 30 days with NAICS code 238210 (electrical contractors) in Stamford, Norwalk, Bridgeport or Danbury. Active filings only, max 100 results. Give me a table of name, registration date, NAICS description and city.
A well-behaved agent maps that onto the tool's inputs:
{
"registeredWithinDays": 30,
"naicsCodes": ["238210"],
"cities": ["Stamford", "Norwalk", "Bridgeport", "Danbury"],
"activeOnly": true,
"maxResults": 100
}
Check the input it chose before you approve the run. If it guessed a field name, fetch-actor-details will correct it on the next try.
Prompts that work well
- "How many new construction companies (NAICS 23) registered in Connecticut last week, by 6-digit code?" The agent pulls the rows and counts them for you.
- "New restaurants (NAICS 722) in New Haven this month, as a CSV."
- "Compare new remodeling companies (236118) in Hartford vs New Haven over the last 60 days."
Why NAICS beats name keywords
Name searches miss companies like "JMK Holdings LLC" that are really electricians. A self-reported NAICS code isn't perfect either, since some filers pick a broad code, but it catches far more of the right businesses than name like '%ELECTRIC%' does.
Prefer code? Here's the raw query
import requests
rows = requests.get("https://data.ct.gov/resource/n7gp-d28j.json", params={
"$select": "name,business_type,date_registration,naics_code,billingcity",
"$where": "date_registration >= '2026-10-01T00:00:00' AND contains(naics_code, '(23')",
"$limit": 5000}, timeout=120).json()
print(len(rows))
naics_code is text like Residential Remodelers (236118), which is why the filter looks for (23. The raw data also includes business email and owner-demographic fields. Leave those out of any marketing list.
The MCP setup uses my tool, Connecticut New Business Leads (paid per result, business-only rows, no emails or personal names). Disclosure: it's mine. The raw query does the same job for free.
Top comments (0)