The loadout MCP server directory is a reviewed catalog of MCP servers for AI coding tools. An MCP server gives an agent a defined connection to an external system: a database, a repository service, an API, a browser workflow, or another tool with a documented interface. Each listing is reviewed before it appears in this directory. The catalog is a place to inspect what a server does, where its source lives, and how it is meant to be installed before it reaches a development machine. See How to sync skills and MCP across machines.
MCP servers
Nothing matches that
No listing in the catalog mentions “”. Try a tool name, a tag, or the author's handle.
About MCP servers
Listings are organized for developers who work across more than one coding environment. Loadout detects the installed harnesses on a paired machine, including Claude Code, Codex, Cursor, GitHub Copilot, and compatible tools as support is available. An install is recorded against the loadout account and sent to the selected paired machine through the background agent. There is no clone, no copy, and no per-machine ritual. The agent writes the MCP configuration in the format the detected harness expects, then keeps the installed state aligned across the fleet.
That matters when the same developer switches between a laptop, a workstation, a remote environment, or a replacement machine. Instead of finding an old config file, copying a command, and checking whether each tool uses the same schema, select the server once and choose the target machine. Loadout carries the install over its private realtime channel. If another paired machine needs the same server, install it there from the same listing; the catalog remains the source of the install record rather than a collection of untracked local edits.
The directory is also useful before any install. Open a listing to review its repository, publisher, tags, and setup guidance. A reviewed listing is not an arbitrary URL or an opaque package name. It is a catalog entry with enough context to decide whether the server belongs in the development workflow. Some servers are useful for reading data, others for making controlled changes, and others for connecting a coding agent to a service already used by the team. The correct choice depends on the tool, the permissions it receives, and the work it is expected to perform.
Installing through loadout requires a signed-in account and at least one paired machine. Browsing does not. Developers can inspect the directory without an account, then pair a machine when they are ready to install. A listing may also provide a manual install path from its repository for teams that need to review the underlying configuration themselves. The result is a single MCP server directory for discovery, review, installation, and repeatable machine setup.