Propraven Mcp — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited Propraven Mcp (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Model Context Protocol server for PropRaven — gives Claude, ChatGPT, Cursor, and any MCP-compatible agent canonical access to 180M+ US parcels with ownership, valuation, permits, deeds, hazard, and market data.
Status: alpha. Hosted endpoint at `mcp.propraven.com` — August 2026. Until then, run it locally (works today).
Requirements: Node ≥18, a PropRaven API key (pz_…) from https://propraven.com/dashboard.
git clone https://github.com/propraven/propraven-mcp.git
cd propraven-mcp
npm install
npm run buildEdit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):
{
"mcpServers": {
"propraven": {
"command": "node",
"args": ["/absolute/path/to/propraven-mcp/dist/index.js"],
"env": {
"PROPRAVEN_API_KEY": "pz_your_real_key_here"
}
}
}
}Restart Claude Desktop. The 8 PropRaven tools (parcel.lookup, parcel.search, parcel.compare, owner.pierce, hazard.score, valuation.estimate, permits.history, sales.history) appear in the tool list.
Look up parcel 37183:0012345 and tell me the owner.Claude should pick parcel.lookup and return canonical record + owner.
Cursor: install via the MCP marketplace (search "PropRaven") once we publish, or wire manually:
// ~/.cursor/mcp.json
{
"mcpServers": {
"propraven": { "command": "node", "args": ["/abs/path/dist/index.js"], "env": { "PROPRAVEN_API_KEY": "pz_..." } }
}
}ChatGPT: Custom GPT → Actions → connect via OAuth (waiting on mcp.propraven.com hosted endpoint — Aug 2026).
| Tool | Purpose | Backed by |
|---|---|---|
parcel.lookup | Resolve one parcel by composite ID, address, or APN | GET /api/v1/parcels/{id} or GET /api/v1/search/full |
parcel.search | Filter parcels by geography + attributes | GET /api/v1/search/full |
parcel.compare | Fetch reports for 2–25 parcels for agent-side ranking | GET /api/v1/parcels/{id}/report × N |
owner.pierce | Resolve a name/entity to its full portfolio | GET /api/v1/owners/{name}/portfolio |
hazard.score | Flood / fire / seismic / windstorm / air-quality / crime composite | GET /api/v1/parcels/{id}/risks |
valuation.estimate | AVM + assessed / market values for a parcel | GET /api/v1/parcels/{id} |
permits.history | Permit timeline for a parcel | GET /api/v1/parcels/{id}/permits |
sales.history | Deed timeline ± UCC liens | GET /api/v1/parcels/{id}/deeds (+ /report if liens) |
Tool descriptions are optimized for agent reasoning — they explicitly call out when not to use each tool, which is the biggest determinant of agent selection accuracy.
The server reads:
| Variable | Default | Purpose |
|---|---|---|
PROPRAVEN_API_KEY | (required) | Bearer token, format pz_... |
PROPRAVEN_BASE_URL | https://api.propraven.com | API host. Use https://propzilla.vercel.app during the DNS cutover window. |
PROPRAVEN_TIMEOUT_MS | 30000 | Per-request timeout in ms. |
# Live-reloading dev (TypeScript, no build step):
PROPRAVEN_API_KEY=pz_test... npm run dev
# Or open the MCP Inspector UI:
PROPRAVEN_API_KEY=pz_test... npm run inspectThe inspector at https://modelcontextprotocol.io/legacy/tools/inspector lets you call each tool manually and see request/response payloads.
Every Claude / ChatGPT / Cursor workflow that touches property data needs an authoritative parcel-lookup tool. There is no canonical property-data MCP today. PropRaven aims to be it before the first-mover window closes (~12–18 months).
Apache-2.0
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.