Overview
Claude (by Anthropic) natively supports the Model Context Protocol. Connect Claude to any ezForge Ready MCP server to let it query business hours, browse offerings, check availability, and more — using its built-in MCP client.Prerequisites
- An ezForge Ready business with a deployed MCP server
- Your MCP server URL (format:
https://<slug>.mcp.ezforge.ai) - Claude Desktop ≥ 0.10 or access to the Claude API with MCP tool-use support
Connecting via Claude Desktop
Add the server to your Claude Desktop configuration file (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):
<slug> with your business slug (visible in the Ready dashboard under Server → Settings).
Claude Desktop will prompt you to authenticate via OAuth 2.1 on first connection.
Connecting via the Claude API
Use the MCP connector in the Claude API to pass tool results from an ezForge Ready server:Maintenance: The
betas=["mcp-client-2025-04-04"] header enables the MCP client beta in the Anthropic API. If the beta graduates to general availability or the beta string date changes, this value must be updated (or removed). Check the Anthropic changelog for the latest status.OAuth 2.1 Authentication
ezForge Ready servers use OAuth 2.1 with PKCE (RFC 7636). Claude handles the full OAuth flow automatically — no manual token management required. Discovery endpoints:Available Tools
All Ready MCP servers expose a common base tool set plus vertical-specific tools:Base tools (all verticals)
Restaurant tools
Salon & Spa tools
Home Services tools
Discovering Businesses
Use the ezForge directory API to find Ready-powered businesses:mcpServerUrl you can pass directly to Claude.
Troubleshooting
Authentication loop — ensure your Claude Desktop is up to date; OAuth PKCE support requires ≥ 0.10. Tool not found — confirm the business has the relevant vertical template deployed. Theget_business_info tool is available on every server; vertical tools depend on the template type.
Rate limiting — the /.well-known/mcp-server-metadata endpoint is rate-limited per IP. If you are building a crawler or aggregator, cache the metadata response for at least 5 minutes.