NEWCloud agents are live.See how

Tooling

MCP

Also called Model Context Protocol

The Model Context Protocol: an open standard for connecting AI applications to outside tools and data (a database, a design file, an issue tracker) through servers that expose a uniform interface. Introduced by Anthropic in 2024 and now supported broadly, it means an integration written once works in any MCP-aware client.

At a glance

Stands for
Model Context Protocol
Introduced
By Anthropic, November 2024
Message format
JSON-RPC 2.0
Roles
Host, client, server
Server features
Tools, resources, prompts

How it works

MCP has three roles. The host is the AI application, such as an editor or a coding agent. Inside it, a client holds a connection to one server. The server wraps some system (GitHub, Postgres, a browser, an internal API) and describes what it offers. Messages are JSON-RPC 2.0, carried over standard input and output for a local process or over HTTP for a remote one.

A server can expose tools, which the model can call to take an action; resources, which are data the application can read into context; and prompts, which are reusable templates a user can invoke. When a session starts, the client asks each server for its tools, and those names and descriptions are given to the model alongside your request.

Why it matters

Before MCP, every integration was written separately for every AI application. With it, a team can write one server for its issue tracker or staging database, and any MCP-aware client can use it. For an agent, that turns "paste the error from the dashboard" into a tool call that fetches it directly.

The cost is context and trust. Every connected tool's description rides along on every request, and a server acts with your credentials, so text it returns (a web page, an issue body) can carry prompt injection.

MCP versus a plain API or a skill

An API is the underlying service; MCP is a standard wrapper that makes it discoverable and callable by a model. If an agent can already run a CLI such as gh in its shell, an MCP server for the same service may add little. A skill packages instructions and scripts that teach an agent how to do a task, while MCP gives it a live connection to a system.

Example

A tool call from client to server (JSON-RPC)
{
  "jsonrpc": "2.0",
  "id": 7,
  "method": "tools/call",
  "params": {
    "name": "search_issues",
    "arguments": { "query": "login timeout", "state": "open" }
  }
}

Common mistakes

  1. Connecting every available server, so tool descriptions crowd the context and the model picks the wrong tool more often.
  2. Granting a server broad write scopes when the task only needs to read.
  3. Trusting text a server returns as if it were your own instructions, which opens the door to prompt injection.

In Fredrin

Fredrin runs a hosted MCP server, so Claude Code, Cursor or any MCP client can create and manage tickets on your board.

Fredrin MCP server

Frequently asked questions

Who created MCP?

Anthropic introduced the Model Context Protocol as an open standard in November 2024, and it has since been adopted by many other AI applications and tool vendors.

Is MCP the same as function calling?

No. Function calling is a model's ability to emit a structured tool request; MCP is the protocol that lets an application discover tools from external servers and route those requests to them.

Do I need MCP to give an agent tools?

No. Agents already have built-in tools such as file editing and a shell, and many services are reachable through a CLI. MCP is useful when you want a reusable connection that works across clients.

Sources

Know the words. Run the work.