io.github.jbechtel-97/dealflowpro

DealFlowPro MCP Server

Analyze multifamily real estate deals from Claude Code, Cursor, Claude Desktop, or any MCP-compatible client.

DealFlowPro is an AI-powered multifamily deal analysis platform. This MCP server wraps the DealFlowPro REST API, giving AI agents access to institutional-grade underwriting calculations.

Demo

What You Can Do

Ask Claude naturally and it calls the right tool:

• "Analyze this 24-unit deal in Charlotte — asking $2M, $13.6K/mo rent, $5.5K expenses"
• "Score this deal: 12 units, $1.5M, $9K monthly income"
• "What's the max I should offer on a property with $28.8K/mo rent if I want 8% cash-on-cash?"
• "Is this property in a flood zone? 2909 Burgess Dr, Charlotte, NC"

Tools

Tool	Description
analyze_deal	Full deal analysis: cap rate, DSCR, cash-on-cash, IRR, DFP Score (0-100), max offer price, yearly cashflows
score_deal	Quick screening: DFP Score + key metrics
reverse_calc	Max offer price from target returns (cap rate, CoC, DSCR, IRR)
market_data	Flood zone, neighborhood income vs state median, job growth for any address

All four carry MCP 2025-06-18 annotations (title, readOnlyHint, destructiveHint, idempotentHint, openWorldHint). Every tool is read-only.

Install as a Claude plugin

This repo doubles as a Claude Code / Cowork plugin. The plugin connects to the remote server at https://dealflowpro.io/mcp rather than the npm package above, which means:

• Eight tools instead of four. Adds list_my_deals, get_deal_analysis, get_my_criteria, and portfolio_summary — your saved deals, buy-box criteria, and pipeline, scoped to your account.
• OAuth instead of an API key. No DFP_API_KEY to export; you sign in to your DealFlowPro account when the plugin connects.

claude --plugin-dir /path/to/dealflowpro-mcp-server

Use the npm package below instead if you want a local stdio server, prefer API key auth, or are wiring DealFlowPro into a non-Claude MCP client.

For Institutional AI Teams

DealFlowPro is built to slot into AI eval pipelines and production AI stacks. The MCP server (this package, plus the remote endpoint at https://dealflowpro.io/mcp) wraps the same engine as the REST API — schema-strict, idempotent, predictable.

Auth + scoping. Every tool call is authenticated, with a Bearer API key or, on the remote endpoint, an OAuth sign-in. Calls are scoped to the account that issued the key or signed in — no cross-tenant access path. Per-tool audit lines land in your account's logs/mcp_tool_calls.log capturing tool name + flattened arg keys (not values) — usage is auditable without exposing deal contents.

Rate limits per tier.

Tier	Daily limit	Monthly	Cost
Free	5 requests, lifetime	—	free, no card
Pay-as-you-go	balance-based	per credits	$1/request
Essentials	50 req/day	~1,500	$79/mo
Premium	200 req/day	~6,000	$149/mo
Enterprise	1,000 req/day	~30,000	$399/mo

Rate limit headers (X-RateLimit-Limit, X-RateLimit-Remaining, X-RateLimit-Reset) returned on every response. 429 responses include Retry-After.

Error codes. Stable, documented at dealflowpro.io/api/docs/#errors — all errors are JSON {success: false, error: {code, message}} shape. Codes: invalid_input (400), invalid_json (400), unauthorized (401), tier_required (403), method_not_allowed (405), payload_too_large (413), rate_limit_exceeded (429), internal_error (500).

Tool schemas. Each MCP tool has a schema-strict input definition. Call tools/list against the remote endpoint to get the canonical schemas at runtime; the REST OpenAPI spec covers the same shapes for the four compute tools (analyze_deal, score_deal, reverse_calc, market_data). The four saved-pipeline tools on the remote server (list_my_deals, get_deal_analysis, get_my_criteria, portfolio_summary) are MCP-only and need an Essentials plan or higher.

Eval harness pattern. For benchmarking, the recommended pattern: maintain a fixture set of (deal payload → expected metrics) pairs, run each against score_deal (the lightest tool; each call still counts as one request, so budget for it under the tier table above), assert metrics within tolerance. The endpoint is deterministic — same inputs produce the same outputs.

Data handling. Request payloads are processed in memory and not persisted to disk. TLS 1.2+ in transit. Anthropic API calls (when DealFlowPro internally uses Claude for document extraction) flow through DealFlowPro's zero-data-retention Anthropic workspace. Full posture: dealflowpro.io/security#api-mcp-data-handling.

Setup

You have two install paths — pick the one that matches your client.

Path A — Remote MCP URL (no local install)

Use this for claude.ai web (Custom Connectors), Claude Code with HTTP transport, or Claude API mcp_servers. The endpoint is https://dealflowpro.io/mcp (Streamable HTTP, MCP 2025-06-18).

claude.ai web / Claude Desktop: DealFlowPro is listed in the Claude Connectors Directory, where you can add it directly, or you can add it yourself as a custom connector — Customize → Connectors → "Add custom connector" → name it DealFlowPro, paste https://dealflowpro.io/mcp → Add → Connect → sign in to DealFlowPro → click Allow. OAuth handles auth automatically (no API key to manage). Then in a chat, click "+" → Connectors → toggle DealFlowPro on.

Add DealFlowPro to Claude → — opens the dialog with the name and URL pre-filled. You still review and confirm before anything is added.

Team / Enterprise plans: members cannot add custom connectors themselves. An organization Owner adds it once under Organization settings → Connectors → Add → Custom (type: Web), and each member then connects their own DealFlowPro account from Customize → Connectors.

Claude Code (HTTP):

claude mcp add dealflowpro \
  --transport http \
  --url https://dealflowpro.io/mcp \
  --header "Authorization: Bearer dfp_sk_your_key_here"

Claude API:

POST https://api.anthropic.com/v1/messages
anthropic-beta: mcp-client-2025-11-20

{
  "mcp_servers": [{
    "type": "url",
    "url": "https://dealflowpro.io/mcp",
    "name": "dealflowpro",
    "authorization_token": "dfp_sk_your_key_here"
  }],
  ...
}

Get a Bearer key (free 5-request key, no card, or pay-as-you-go credits) at dealflowpro.io/api.

Full setup details + Connected Apps revocation: dealflowpro.io/api/docs#mcp.

Path B — Local stdio (this npm package)

Use this anywhere you want the MCP server running locally instead of remote — or for a client that has no remote-URL path of its own. Claude Desktop supports both: remote via Settings → Connectors, or the local config below.

1. Get an API key. Visit dealflowpro.io/api to claim a free 5-request key (no card) or buy pay-as-you-go credits. Delivered by email.

2. Add to your client.

Claude Desktop — claude_desktop_config.json:

{
  "mcpServers": {
    "dealflowpro": {
      "command": "npx",
      "args": ["-y", "dealflowpro-mcp"],
      "env": {
        "DFP_API_KEY": "dfp_sk_your_key_here"
      }
    }
  }
}

Claude Code:

claude mcp add dealflowpro -e DFP_API_KEY=dfp_sk_your_key_here -- npx -y dealflowpro-mcp

Use It

Just ask Claude about a deal. It automatically picks the right tool.

Example Output

## Deal Analysis Results

**DFP Score: 19/100 (Poor)**

### Key Metrics
| Metric | Value |
|--------|-------|
| Cap Rate | 4.48% |
| DSCR | 0.86 |
| Cash-on-Cash | -1.95% |
| IRR | -12.17% |
| NOI | $89,524 |

### Max Offer Price
**$1,197,476** (binding constraint: cash-on-cash)

API Documentation

Full REST API docs at dealflowpro.io/api/docs

Pricing

• Pay-as-you-go: $1/API request, starting at $5
• Essentials: 50 API requests/day included ($79/mo)
• Premium: 200 requests/day + reverse calculator ($149/mo)
• Enterprise: 1,000 requests/day ($399/mo)

Requirements

• Node.js 18+
• DealFlowPro API key

About DealFlowPro

DealFlowPro automates multifamily underwriting for 2-200 unit properties. Upload a broker email or OM and get a full investment analysis in 10 minutes instead of 2 hours. Used by operators, syndicators, and PE firms to screen deals faster.

dealflowpro.io

License

MIT