For AI agents
Boardwalk for AI agents
Boardwalk ships a live MCP server so AI assistants can find US real estate development projects 12–24 months before a permit or bid exists — at the planning, zoning, and entitlement stage. Ask Claude, ChatGPT, Cursor, or Windsurf for pre-construction projects in any city, and the assistant queries Boardwalk's 1,158,457 indexed projects across all 50 states (2026 YTD) directly.
What is the Boardwalk MCP server?
The Boardwalk MCP server (https://mcp.boardwalkai.com/mcp) is a remote, OAuth-secured tool endpoint any Model Context Protocol client can connect to. Once connected, an assistant can search Boardwalk's land-use intelligence database — development projects extracted from city and county meeting records nationwide — without writing code or leaving the chat. Each record is a named project with an exact address, magnitude, entitlement stage, developer and applicant parties, and the meeting date it surfaced. The server exposes read-only tools and uses OAuth 2.1 PKCE with Dynamic Client Registration.
Which tools does it expose?
describe_data
What the dataset can and cannot answer: groupable fields, metrics with their denominators, coverage, and the questions this data does not support. Free.
list_locations
States, counties and cities with the numeric IDs that geographic filters require. Free.
get_taxonomy
The land-use taxonomy: property types, subtypes, asset classes, action types. Free.
count_projects
How many projects match a filter set, and what fetching them would cost. Free — call it before spending anything.
analyze_projects
Counts and distributions grouped by city, county, type, construction type or month. Free, and returns no project identity.
rank_jurisdictions
Leaderboards of cities, counties or states by a metric, with a minimum denominator and a comparison baseline. Free.
get_projects
Project records matching a filter set. 1 credit per project returned; an explicit limit is required.
get_project
One project in full detail, including its source-document references. 1 credit.
get_document
The source meeting document a project was extracted from. Standard record rate.
ai_search_projects
Natural-language search via semantic + structured retrieval. 1 credit per project returned.
get_ai_search_result
Collect a submitted AI search. Free — the records were billed when the job ran.
get_credit_balance
The connected account's remaining credits. Free. Read-only.
Try it: connect in your terminal
# Boardwalk MCP server (remote, streamable HTTP + OAuth) claude mcp add --transport http boardwalk https://mcp.boardwalkai.com/mcp
Frequently asked questions
Can an AI assistant use Boardwalk without writing code?
Yes. In Claude, add the connector at boardwalkai.com/connect/claude/ and sign in once; the assistant then invokes Boardwalk tools automatically. In ChatGPT, Cursor, or Windsurf, point the client at the same MCP server or use the API key path in the skill files.
What does it cost to call Boardwalk?
Every project record costs 1 credit, whether it came from structured search or AI search. Taxonomy and location lookups are free.
How fresh is the Boardwalk data?
Boardwalk recomputes rolling weekly windows and refreshes nationwide coverage every week. The public index currently shows 1,158,457 U.S. development projects year-to-date in 2026. Data is refreshed weekly through 2026-08-16.