"""HTTP transport entry point for SF Permits MCP server.
Exposes the same 27 tools as the stdio server, but over Streamable HTTP
for Claude.ai custom connector access.
Uses mcp[cli] package (same as Chief MCP server — proven claude.ai compatibility).
Run locally:
python -m src.mcp_http
Deploy:
Railway service with Dockerfile.mcp
Connect from Claude.ai:
Settings > Integrations > Add MCP server > paste URL + /mcp
"""
import os
from mcp.server.fastmcp import FastMCP
# Import all tool functions (same as server.py)
from src.tools.search_permits import search_permits
from src.tools.get_permit_details import get_permit_details
from src.tools.permit_stats import permit_stats
from src.tools.search_businesses import search_businesses
from src.tools.property_lookup import property_lookup
from src.tools.search_complaints import search_complaints
from src.tools.search_violations import search_violations
from src.tools.search_inspections import search_inspections
from src.tools.search_entity import search_entity
from src.tools.entity_network import entity_network
from src.tools.network_anomalies import network_anomalies
from src.tools.predict_permits import predict_permits
from src.tools.estimate_timeline import estimate_timeline
from src.tools.estimate_fees import estimate_fees
from src.tools.required_documents import required_documents
from src.tools.revision_risk import revision_risk
from src.tools.validate_plans import validate_plans
from src.tools.analyze_plans import analyze_plans
from src.tools.recommend_consultants import recommend_consultants
from src.tools.permit_lookup import permit_lookup
from src.tools.search_addenda import search_addenda
from src.tools.list_feedback import list_feedback
# Phase 7 tools (project intelligence)
from src.tools.project_intel import run_query, read_source, search_source, schema_info, list_tests
# ── Create MCP server with HTTP transport config ──────────────────
port = int(os.environ.get("PORT", 8001))
mcp = FastMCP(
"SF Permits",
instructions=(
"SF Permits MCP server — query San Francisco public permitting data. "
"27 tools across 7 phases: live SODA API queries, entity network analysis, "
"permit decision tools, plan set validation, AI vision analysis, "
"addenda routing search across 3.9M+ records, and project intelligence "
"tools for read-only database queries, source code reading, codebase search, "
"schema introspection, and test inventory."
),
host="0.0.0.0",
port=port,
stateless_http=True,
streamable_http_path="/mcp",
)
# Register all 27 tools
mcp.tool()(search_permits)
mcp.tool()(get_permit_details)
mcp.tool()(permit_stats)
mcp.tool()(search_businesses)
mcp.tool()(property_lookup)
mcp.tool()(search_complaints)
mcp.tool()(search_violations)
mcp.tool()(search_inspections)
mcp.tool()(search_entity)
mcp.tool()(entity_network)
mcp.tool()(network_anomalies)
mcp.tool()(predict_permits)
mcp.tool()(estimate_timeline)
mcp.tool()(estimate_fees)
mcp.tool()(required_documents)
mcp.tool()(revision_risk)
mcp.tool()(validate_plans)
mcp.tool()(analyze_plans)
mcp.tool()(recommend_consultants)
mcp.tool()(permit_lookup)
mcp.tool()(search_addenda)
mcp.tool()(list_feedback)
# Phase 7 tools (project intelligence)
mcp.tool()(run_query)
mcp.tool()(read_source)
mcp.tool()(search_source)
mcp.tool()(schema_info)
mcp.tool()(list_tests)
# ── Health check endpoint (for Railway) ───────────────────────────
from starlette.requests import Request as StarletteRequest
from starlette.responses import JSONResponse
from starlette.routing import Route
async def health_check(request: StarletteRequest) -> JSONResponse:
return JSONResponse({
"status": "healthy",
"server": "SF Permits MCP",
"tools": 27,
})
mcp._custom_starlette_routes.append(Route("/health", health_check))
mcp._custom_starlette_routes.append(Route("/", health_check))
if __name__ == "__main__":
mcp.run(transport="streamable-http")
