Start here

What is MCP?

MCP (Model Context Protocol) is a standard way for AI agents and coding tools to connect to external services through structured tool servers.

In practice, MCP gives your agent a list of tools (with schemas) it can call - like "search docs", "query a database", or "create an issue" - without you writing a custom integration for each client.

MCP client

Your agent or IDE

Codex, OpenCode, VS Code, Cursor, Claude, and others.

MCP server

A tool server endpoint

It exposes tools and schemas over JSON-RPC.

MCP tool

A callable function

The agent calls it with validated arguments.

Why people get stuck

MCP is simple at one tool. It gets messy when you have many tools across multiple clients, projects, and machines.

  • Each client has a different config format.
  • Auth differs per provider (OAuth vs API keys vs tokens).
  • You end up duplicating setup everywhere.
  • As tool count grows, the agent context surface gets noisy fast.

Where Loadoutz fits

Loadoutz is the control plane + gateway for MCP.

  • In the web app, you manage a workspace (your tool library), credentials, and loadouts (curated tool sets).
  • A connection creates one stable owner URL. Your client connects once to that URL and completes OAuth.
  • Switching loadouts changes which tools are exposed without rewriting client config.
  • Code mode gives the agent a multi-step execution lane for jobs that need several tool calls or lightweight scripting in one run.

Need help next?