Anthropic introduced MCP in November 2024, and it has since been adopted by many AI clients. An MCP server describes the tools it offers, such as search leads or create campaign, with their inputs. The AI client reads that list and calls a tool when your request needs it, then shows you the result.
How a connection works
- You add the server's URL to your AI client.
- You sign in (OAuth) or paste an API key, which decides what the assistant may do.
- You ask in plain language, and the client picks the tools to call.
"Find 50 marketing directors at US fintech companies, put them in a new table, and draft a two step campaign. Do not launch it."
What MCP means for outbound
With MCP, the assistant you already use becomes a front end for your sales tools. You can build a list, check mailbox health or read replies from a chat window, without switching apps. The key question is permission: a good MCP server lets you give a read only connection, limits actions to what your role allows, and lets you revoke access in one click.
ZicReach runs an MCP server at app.zicreach.com/api/mcp over Streamable HTTP, with guides for 24 clients including Claude, ChatGPT, Cursor, VS Code and n8n. Claude, ChatGPT (developer mode), Mistral Le Chat and Claude Code can sign in with OAuth; others use a read only or read and write API key. See MCP and API, and ZicAgent for the same abilities inside the app.