Model Context Protocol

Also called: MCP, MCP server.

MCP was introduced by Anthropic in November 2024 as an open protocol. An application exposes an MCP server with a set of tools (functions with descriptions and typed inputs) and resources. An AI client connects to it, reads what the tools do, and calls them when a user's request needs them.

Before MCP, each assistant needed its own plugin format. With MCP, one server works with any client that supports the protocol, which now includes many AI assistants, coding tools and agent frameworks.

For lead generation, MCP means you can ask an assistant for leads in plain language ("find plumbers in Austin with a verified email") and it calls the lead tool directly, starts a search, checks progress, and reads results back into the conversation or into a next step such as drafting emails.

Treat MCP access like API access. A connected assistant can spend credits and read data on your behalf, so use scoped API keys and review what it does.

Example

In Claude, a user asks for HVAC contractors in Phoenix. The assistant calls start_search, polls search_status, then calls find_leads and summarizes the businesses with verified emails.

In Leadraft

Leadraft runs an MCP server with six tools: find_leads, get_lead, start_search, search_status, list_searches and credits. See the MCP docs.