publicmcp

publicmcp

Build a PUBLICMCP-compliant business identity MCP server in minutes.

PUBLICMCP is an open standard for public-facing business identity MCP servers — structured, AI-readable business data that any AI agent or MCP client can query. This package is the reference implementation.

Full spec and registry: github.com/joeprovence/publicmcp

Quickstart

pip install publicmcp
publicmcp init          # create business.json
publicmcp serve         # start server on :8000

Point your MCP client at http://localhost:8000/mcp. That's it.

What it does

publicmcp serve starts a FastMCP server that exposes your business.json as structured MCP tools. AI agents (Claude, Gemini, GPT, Continue, etc.) can call these tools to learn about your business, qualify a prospect, or pull your service and portfolio data — without hallucinating.

Browsers hitting the root URL are redirected to your publicmcp_page landing page. AI agents get MCP.

CLI reference

publicmcp init [--output business.json] [--force]
    Create a starter business.json template.

publicmcp serve [--business business.json] [--host 0.0.0.0] [--port 8000] [--domain your.domain.com]
    Start the server. --domain sets the URL in the discovery endpoint.

Endpoints

Path	Method	Description
/mcp	POST	MCP (streamable-http) — for AI agents
/.well-known/publicmcp.json	GET	Discovery endpoint
/info	GET	Plain-text server summary
/	GET	Discovery payload (JSON)

Tools (PUBLICMCP v0.2)

All tools are read-only. Data comes from your business.json — no external calls.

Tool	Required	Description
get_info	✓	Name, tagline, description, contact, founders, hours, social profiles
get_services	✓	Services with pricing, highlights, and timeline
get_location	✓	Offices, service areas, headquarters, geo
get_portfolio	—	Portfolio items, filterable by market and category
get_sales_context	—	Objection handling, pain points, ideal client profile (requires sales_agent in business.json)
qualify_prospect	—	Fit score, vertical detection, service recommendation (requires primary_markets in business.json)

Plus one prompt: sales_workflow — tool chaining guide for consultative sales AI agents.

business.json

publicmcp init creates a starter template. The minimum required fields for spec compliance:

{
  "name": "Your Business",
  "services": [...],
  "locations": [...]
}

A full v0.2 schema is in the spec.

Discovery

Every server exposes a /.well-known/publicmcp.json endpoint so AI agents can auto-discover capabilities:

{
  "publicmcp_version": "0.2",
  "auth": "none",
  "endpoint": "https://mcp.yourdomain.com/mcp",
  "tools": [...]
}

Production deployment

For production, run behind a reverse proxy (Caddy or nginx) with SSL:

# Caddyfile
mcp.yourdomain.com {
    reverse_proxy 127.0.0.1:8000
}

Then run with your public domain:

publicmcp serve --business business.json --host 127.0.0.1 --port 8000 --domain mcp.yourdomain.com

Registry

Live deployments are listed in the PUBLICMCP registry. 15 servers active across the US and Canada as of April 2026.

To register your server, open a PR to registry/servers.yaml.

Links

• Spec: SPEC.md
• Registry: registry/servers.yaml
• Standard: publicmcp.org
• PyPI: pypi.org/project/publicmcp

License

Apache 2.0 — see LICENSE.