Skip to main content
The Model Context Protocol (MCP) is an open protocol released by Anthropic for AI models to access external tools, data, and systems in a standardized way. It’s based on JSON-RPC 2.0 and uses STDIO or HTTP + SSE transports.

​
Core concepts

Server

A process that exposes tools (tools), resources (resources), and prompts (prompts).

Client

Claude Desktop, Cursor, custom agents — attach to a Server and call tools.

Tool

A callable function with a name, description, and input schema (JSON Schema).

Transport

Two standard transports: STDIO (local process) or Streamable HTTP (remote).

​
Standard methods

The MCP methods the Bridger Gateway currently handles are as follows.
Bridger is a tools-focused gateway. resources/* and prompts/* are standard MCP methods but are not handled by the Gateway today; any other method responds with Method not found.

​
Bridger and MCP

Bridger is an MCP Server hosting platform. Without writing, deploying, or operating your own MCP server, any OpenAPI spec is instantly converted into MCP tools.
  • A single domain (mcp.datari.kr) multiplexes hundreds of tools by name (getultrashortcast, getapttrade…). Official preset servers are distinguished by {category}/{name} IDs, while BYOAPI tools use the @{owner}/{slug} namespace.
  • Use all registered tools with a single API Key — no need to spin up a separate Server.
  • Clients work as-is with any MCP-compatible client (Claude Desktop, Cursor, a ChatGPT Actions adapter, your own SDK).

​
Next steps

View the architecture

How the Gateway, Registry, and middleware fit together.

Call your first tool

Run a real JSON-RPC call yourself.