Introduction
NT lets you build AI agents by describing them in a simple text file, instead of writing code.
Instead of writing code, you describe what you want in a plain text file (a file
that ends in .nt). Then you run one command, and NT turns your description into
a working agent that talks to a real AI model.
Think of it like a recipe. You list the ingredients (the model, the tools, the steps), and NT cooks it for you.
Why NT
Building an AI agent normally means writing a lot of code to glue things together: the model, the tools it can use, the files it can read, how it hands work to other agents, and so on. That is a lot of plumbing.
NT replaces the plumbing with a short, readable file. You write what the agent should do in plain words, and NT wires everything up.
You do not need to be a programmer
If you can write a shopping list, you can write an .nt file. It uses simple name: value lines
and short lists. No programming needed.
What you can describe
An .nt file can contain any of these building blocks:
Agent
The main worker. It reads your request and does the job.
Tools
Extra powers, like running a command or calling a website.
Skills
Reusable instructions you can share between agents.
Sandboxes
A safe space where the agent can read and write files.
Subagents
Helper agents your main agent can hand work to.
Workflows
A list of steps that run one after another.
A tiny example
Here is a complete agent. It estimates a person's age from clues:
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: stringThat is the whole thing. Run it, and you have a working agent.
How you use it
Write a file
Describe your agent in an .nt file.
Run one command
Type nt run and NT brings your agent to life against a real AI model.
Get an answer
Give it some input, and it replies. That is it.