munimind-mcp 0.3.1

MCP server for MuniMind — 2,500+ NYC civic datasets with tools + prompts + resources for Claude Desktop, Claude Code, Cursor.

munimind-mcp

MCP server for MuniMind — query 2,500+ NYC civic datasets directly from Claude Desktop, Claude Code, or Cursor. More cities shipping soon on the same schema.

What you can do

Once installed, ask Claude natural-language questions that route through MuniMind's API:

• "List all MuniMind datasets about housing violations."
• "Show me the schema for NYC's HPD violations dataset."
• "How many DOB violations were issued in Queens in 2024?" (runs SQL)
• "What's the ownership history for BBL 1000010010?" (Answers API)
• "Preview the first 20 rows of the NYC taxi zones dataset."
• "/analyze_property 1000010010" (curated prompt — chains multiple tools)

Install

pip install munimind-mcp

Or via uv (recommended — MCP clients use uvx):

uvx munimind-mcp

Get an API key

1. Go to https://munimind.com/developers
2. Sign up — first 5,000 credits are free, no credit card.
3. Create an API key in the developer dashboard
4. Copy the key (starts with mm_live_)

Configure your MCP client

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "munimind": {
      "command": "uvx",
      "args": ["munimind-mcp"],
      "env": {
        "MUNIMIND_API_KEY": "mm_live_..."
      }
    }
  }
}

Restart Claude Desktop.

Claude Code

claude mcp add munimind uvx munimind-mcp -e MUNIMIND_API_KEY=mm_live_...

Cursor

Add to .cursor/mcp.json in your workspace or ~/.cursor/mcp.json globally:

{
  "mcpServers": {
    "munimind": {
      "command": "uvx",
      "args": ["munimind-mcp"],
      "env": { "MUNIMIND_API_KEY": "mm_live_..." }
    }
  }
}

Auto-discovery

Any client that supports the 2026 .well-known/mcp.json spec can auto-discover MuniMind via:

curl https://api.munimind.com/.well-known/mcp.json

What's included (v0.3)

7 tools

Tool	Description	Credit cost
list_datasets	Browse the public catalog (optionally filter by city, keyword)	0
describe_dataset	Full metadata + schema for one dataset	0
preview_dataset	First 100 rows of a dataset	0
query_dataset	Run SQL against DuckLake-backed datasets (supports time-travel + Parquet)	dynamic (~1 credit / 10 MB scanned)
property_summary	Per-property summary via the Answers API	1
property_intelligence	Full property dossier — risk, flood, environmental, ownership	5
owner_portfolio	Cross-property analysis for an owner / LLC	10

6 curated prompts

Claude Desktop surfaces these in its slash-menu. Each prompt is a multi-step workflow that chains tools end-to-end.

• /analyze_property — full diligence on one property
• /investigate_owner — map a portfolio, detect shell-LLC links
• /area_development_scan — permits + BSA/LPC/CPC cases + lobbying in an area
• /meeting_digest — summarize CPC/BSA/LPC/council activity
• /precedent_search — find similar BSA/LPC/CPC decisions
• /suggest_sql — turn a plain-English question into a DuckLake SQL query (NL2SQL via your client's LLM — zero server-side inference cost)

4 resources

MCP resources are context your client reads without spending a tool call.

• munimind://catalog — full dataset catalog
• munimind://schemas/index — every queryable schema
• munimind://schemas/{slug} — column-level schema for a dataset
• munimind://city/{city}/stats — per-city rollup

Environment variables

Name	Required	Default	Notes
MUNIMIND_API_KEY	yes	—	Your API key from the developer dashboard
MUNIMIND_API_BASE_URL	no	https://api.munimind.com	Override for staging / self-hosted

License

MIT

Links

• Docs: https://munimind.com/developers/mcp
• Interactive API reference: https://munimind.com/developers/reference
• In-browser SQL playground: https://munimind.com/developers/playground
• Issues: https://github.com/munimind/munimind-mcp/issues