The operating system for
commercial real estate

AI-native. MCP-native. 146 agent tools. DeepBreadth isn't a CRM with AI bolted on — the agent is the primary interface, and every workflow was designed for it from day one.

Investor OS

96 tools

For acquisition teams, PE funds, and owner-operators across self-storage, industrial, MHP, multifamily, and 7 other CRE asset types.

  • 17-stage deal pipeline with AI underwriting engine
  • Automated financial extraction — T-12s, rent rolls, OMs
  • Conviction scoring, IC workflow & voting
  • Structured outreach sessions & campaign automation
Explore Investor OS

Broker OS

95 tools

For CRE brokers who manage listings, match buyers, and advise owners on portfolio strategy.

  • Listing lifecycle with 100-point buyer match scoring
  • Seller conviction scoring & owner pipeline
  • BOV generation, portfolio advisory & 1031 tracking
  • 4 stackable add-on modules — underwriting, market intel, BOV, portfolio
Explore Broker OS

Need access to both Investor OS and Broker OS? We offer unified accounts for firms that acquire and broker.

Contact Us

Start your 30-day free trial with full access. White-glove onboarding included.

// architecture

Built for the agent era

DeepBreadth exposes 136 tools via a production MCP server. Your AI assistant — Claude, ChatGPT, or any MCP-compatible agent — connects directly to your deal pipeline, CRM, and market data.

Agent-First Interface

The AI agent is the primary user. 146 tools handle prospecting, underwriting, deal management, and outreach. The dashboard is one of many windows into the system.

MCP Server — 136 Tools

Production MCP server with OAuth 2.1, API key, and JWT auth. Claude Desktop, the Claude web app, and ChatGPT connect natively. 55 write operations require human confirmation.

3-Tier Model Routing

Haiku for fast lookups (~70%), Sonnet for synthesis (~25%), Opus for complex reasoning (~5%). Prompt caching cuts inference costs ~46%. Every token is optimized.

Immutable Audit Trail

Every tool call — agent, MCP, or UI — is logged with full I/O, user, org, model, and token count. Confirmation-gated writes. Row-level security at the database engine.

mcp.tools.list()
> Connected to DeepBreadth MCP Server
> 136 tools available // 80 read-only, 55 confirmable, 3 meta
// sample tool calls
search_contacts("MHP owners in Nashville")
run_underwriting_analysis(deal_id) // DCF, IRR, sensitivity
generate_bov(listing_id) // 3-approach valuation
get_conviction_insights() // rising, falling, needs attention
suggest_buyers_for_listing(listing_id) // 100-point match scoring
adjust_conviction_score(owner_id, +20) // requires confirmation

Why MCP-native matters for your firm

The CRE industry is shifting from human-navigated software to agent-driven systems. MCP (Model Context Protocol) is the open standard that lets AI agents use tools — the same way a browser uses URLs. A platform built on MCP doesn't just support AI. It was designed so AI agents can run your pipeline, underwrite your deals, and manage your relationships directly — with human judgment reserved for the decisions that actually require it.

146
Agent tools
Every CRE workflow — from prospecting to close — is an executable tool
136
MCP-exposed
Your external AI agents connect and operate with the same capabilities
55
Human-confirmed
Write operations require your approval — the agent proposes, you decide