What is a cost segregation MCP server?

MCP (Model Context Protocol) is the open standard that lets AI assistants like Claude, Cursor, and ChatGPT call external tools. The Cost Seg Smart MCP server is a hosted endpoint that gives any MCP-compatible assistant two tools: real-time cost segregation study pricing with estimated Year-1 depreciation for a specific US property, and secure Stripe checkout link generation for ordering a study.

Is the Cost Seg Smart MCP server free to use?

Yes. The endpoint is free and requires no API key or account. You only pay if you decide to order a study, and pricing is always confirmed before any checkout link is generated. Engineering-based residential studies start at $495.

Are the numbers it returns tax advice?

No. The quote tool returns the study fee for a property plus estimated Year-1 depreciation and illustrative tax impact based on calibrated reclassification ratios. Actual results depend on the property, the completed engineering-based study, and the taxpayer's situation. Your CPA makes the tax calls.

Where do the prices come from?

The same pricing engine that powers checkout on costsegsmart.com. The MCP tool and the order form read one shared price schedule, so a quote from your AI assistant always matches what you would pay on the site.

For AI-native investors

Cost segregation, callable from your AI stack

A free, hosted MCP server that lets Claude, Cursor, ChatGPT, or any MCP-compatible agent pull real-time study pricing and estimated Year-1 depreciation for a specific US property, mid-conversation, while you are underwriting the deal.

 POST https://costsegsmart.com/mcp 

Streamable HTTP · JSON-RPC 2.0 · no API key · no account · live health check

Listed in the official MCP Registry (com.costsegsmart/cost-seg-smart), on Smithery, and LobeHub. Manifest and docs on GitHub.

Why this exists

If you underwrite rentals in a spreadsheet with an AI assistant open in the next pane, the after-tax picture usually waits until someone emails a cost seg firm for a quote. This server removes that step. Ask your assistant what a study costs on the property you are modeling and what the estimated first-year deduction looks like, and it answers with live numbers from the same pricing engine that runs our checkout. No sales call, no form, no waiting.

CPAs and analysts use it the same way: screen a client list or an acquisition pipeline for properties where a study is clearly worth the fee, before anyone commissions anything.

Two tools, no keys

1. `get_cost_seg_quote` read-only

Takes a property type, purchase price, and optionally your marginal tax bracket. Returns the exact study fee plus estimated Year-1 depreciation, illustrative tax impact, and the ROI multiple on the study fee. Real response from the live endpoint:

// request
{ "property_type": "str", "purchase_price": 750000, "tax_bracket": 0.32 }

// response (live, unedited)
{
  "property_type_label": "Short-Term Rental",
  "purchase_price": 750000,
  "study_cost": 995,
  "deduction_estimate_year_one": 184500,
  "tax_savings_year_one": 59040,
  "roi_multiple": 59,
  "currency": "USD",
  "disclaimer": "Estimate based on industry-standard construction cost
   data, calibrated reclassification ratios, and IRS-aligned methodology.
   Final result varies with property condition, finishes, and
   component-level engineering. Not tax advice."
}

2. `get_cost_seg_payment_link`

After you confirm the quoted fee, generates a Stripe Checkout URL carrying the property details. Payment happens in your browser on Stripe, never inside the chat. The engineering-based study is generated and delivered by email, most residential studies the same day. Assistants are instructed to always confirm the fee with you first.

Add it to your assistant

Claude.ai / Claude Desktop

Settings → Connectors → Add custom connector → paste the endpoint URL. Or in claude_desktop_config.json:

{
  "mcpServers": {
    "cost-seg-smart": {
      "transport": "streamable-http",
      "url": "https://costsegsmart.com/mcp"
    }
  }
}

Claude Code

claude mcp add --transport http cost-seg-smart https://costsegsmart.com/mcp

Cursor

{ "cost-seg-smart": { "url": "https://costsegsmart.com/mcp" } }

ChatGPT and everything else

Any client that speaks MCP over Streamable HTTP works: add a connector pointing at the endpoint URL. There is no auth handshake to configure.

Prompts worth stealing

"I'm buying a $750K short-term rental in Destin. What would a cost seg study cost, and what's the estimated Year-1 deduction at a 32% bracket?"
"Compare the study fee and estimated first-year tax impact for a $400K long-term rental vs a $400K STR."
"Here are the 6 properties in my portfolio spreadsheet. For each one, get the study cost and Year-1 estimate, then rank them by ROI on the study fee."

The fine print, up front

Estimates, not tax advice. The quote tool returns the study fee plus estimated figures from calibrated reclassification ratios. The deliverable that supports a filing is the completed engineering-based study, and your CPA makes the tax calls.
One price schedule. The tool reads the same pricing engine as the order form. A quote in chat always matches the site. Residential studies start at $495.
No data collection. Quote calls need a property type and price, nothing else. No account, no email, no key.
Agent attribution. Clients may send an X-Agent-Name header; it appears in the checkout disclaimer so the buyer always knows an agent prepared the order.

Prefer a human-shaped starting point?

Read a complete sample study, or get a free preliminary depreciation estimate prepared for your property. CPAs and firms: wholesale and white-label options.