NT (agent-lang.xyz)
Language support for NT — a small declarative language (.nt files) for
building AI agents. This extension gives .nt files syntax highlighting and
full IntelliSense: go to definition, hover, completion, outline, and find
references, across imported files.
Features
Syntax highlighting for import statements, declaration keywords (agent,
subagent, sandbox, tool, skill, workflow, provider, config), entity
names, field keys, env(...), block scalars, strings, and numbers.
IntelliSense (works within a file and across imported files):
- Go to Definition — F12 or Ctrl/Cmd-click a tool, subagent, sandbox, skill,
workflow, or agent name to jump to its declaration, in the same file or
another. Works on
import ./path.nt paths too.
- Hover — see an entity's kind, description, and defining location. Built-in
tools (
fs_read, fs_write, fs_list, bash) show a description.
- Completion — inside
tools: / subagents: / skills: lists and after
sandbox:, suggests the matching declared names (plus built-in tools). At the
start of a line, suggests declaration keywords.
- Outline & breadcrumbs — every declaration appears in the Outline view and
the "Go to Symbol" (Ctrl/Cmd-Shift-O) picker.
- Find All References — right-click a name → Find All References.
The index refreshes automatically as you edit, save, create, or delete .nt
files, so definitions stay current across the whole workspace.
Install
In VS Code: open the Extensions view (Ctrl/Cmd-Shift-X), search for
"NT (agent-lang.xyz)", and click Install. Or from a terminal:
code --install-extension KartikMehta.nt-agent-lang
In Cursor, Windsurf, or VSCodium: install from
Open VSX (search
"NT (Agent Lang)" in the Extensions view).
Any file ending in .nt is picked up automatically.
What is NT
NT lets you describe an AI agent — its model, tools, skills, sandbox, subagents,
and typed input/output — in one clean .nt file, then run it with a single
command:
agent age
description: Guesses a person's age from clues.
model: anthropic/claude-sonnet-5
instructions: |
Guess the person's most likely age from the clues you are given.
input:
clues: string
output:
age: number
reason: string
Install the CLI with npm install -g @age.nt/nt and run nt run age. See the
full guide at https://agent-lang.xyz.
