MCP server
A service that implements the Model Context Protocol to expose tools, resources, or prompts that an AI agent can discover and use over a standard interface.
An MCP server is a program that speaks the Model Context Protocol and offers capabilities to a connected AI client. Where the protocol defines the contract, the server is a concrete endpoint that implements it: it advertises a set of tools, resources, or prompts, and responds to the client's calls. A client such as an AI assistant or coding agent connects to the server, lists what it provides, and invokes the pieces it needs.
MCP servers can run locally or be reached over the network. A remote server is typically exposed over HTTP (for example Streamable HTTP), which lets a hosted service offer its capabilities to many clients at once. Each server is responsible for its own auth, its own tool definitions, and the behaviour behind each tool, so the same client can mix capabilities from several independent servers.
Proxygate is a remote MCP server for buying API access. It exposes tools including browse_apis, describe_endpoint, get_pricing, check_balance, call_api, get_usage, and rate_seller. An agent connects once and can then search the catalog, inspect endpoints, pay per request from a prepaid USDC balance, and review its usage, all through this one server.
References
Related concepts
MCP server: frequently asked questions
An MCP server is a service that implements the Model Context Protocol to expose tools, resources, or prompts that an AI agent can discover and use over a standard interface. The client connects, lists what the server offers, and calls what it needs.
Yes. MCP servers can run locally or remotely. A remote server is usually exposed over HTTP, such as Streamable HTTP, so a hosted service can offer its tools to many clients at once.
It lets an agent buy API access. Tools such as browse_apis, describe_endpoint, get_pricing, check_balance, call_api, get_usage, and rate_seller let the agent search the catalog, pay per request from a prepaid balance, and review usage.