The Model Context Protocol (MCP) is a standard way to expose tools, data, and prompts to AI clients like Claude and Cursor. An MCP server is the thing that offers those tools; the AI client is the thing that calls them.
Why it matters for an API
Your REST API is built for developers writing code. An MCP server makes the same capabilities available to an AI assistant in plain language. Instead of a developer writing a fetch call, a user can ask their assistant to "add this person to my contacts" and it happens.
What a good MCP server needs
- Tools: named actions with typed inputs (e.g.
create_contact). - Auth: ideally OAuth, so users approve access instead of pasting keys.
- Scoping: every call limited to the account that authorized it.
Example
ContactAPI exposes its contact operations over MCP at /mcp, secured
with OAuth 2.1 and dynamic client registration. Point a compatible client at the URL
and it registers itself, prompts you to approve, and can then manage your contacts.
If you're building an API in 2026, an MCP server is fast becoming table stakes. It's how your product shows up inside the AI tools people already use.
Where ContactAPI fits
ContactAPI is an open-source REST API for saving and managing contacts. You POST an email plus any JSON fields, get simple CRUD back, and can reach the same data from an AI client over its MCP server. No CRM, no pipeline stages, no lock-in.