Concepts

Code intelligence

LoopCode uses language servers for definitions, references, diagnostics and safe renames across your project — for 15 languages out of the box, and any other you configure.

#What it gives the agent

With a language server, LoopCode does not have to guess from text search:

• Definition and references — where a symbol is defined, through imports, and every use of it: the exact symbol, not every word that matches.
• Callers, callees and implementations — the functions that call it or that it calls, and what implements an interface.
• Hover, outline and symbols — a symbol's type and documentation, a file's symbols with their line ranges, and a symbol found by name across the workspace, without reading whole files.
• Diagnostics — the errors an edit causes, in the edited file and in the files that use it, as soon as it is made.
• Rename — a symbol renamed everywhere it is used, as one reviewed diff applied all at once.

#Supported languages

A server starts when the agent first reads a file in that language. If the server is not on your PATH and can be downloaded, LoopCode fetches it for itself. Set LOOPCODE_LSP_DOWNLOAD=0 to turn downloads off and use only servers you installed.

/lsp in the terminal, or Settings → Agent behavior in the desktop app, shows the language servers in use and their state.

#Add another language

Configure your own servers in ~/.loopcode/settings.json:

{
  "lsp": {
    "elixir": {
      "command": "elixir-ls",
      "extensions": [".ex", ".exs"],
      "rootMarkers": ["mix.exs"]
    }
  }
}

Language servers are read only from your own settings file, never from a repository's: a server starts when the agent reads a file, so a cloned repository must never choose a program to run on your machine.