TrussOpen source

Truss CLI

Automate coding work from your shell.

Run local-model tasks in a terminal, automate repeatable work in scripts, or host the JSON-lines service behind other clients.

View CLI on npmRead documentation
Truss CLI streaming a local-model chat response in a terminal.
Truss CLI in action

What it is for

Purpose-built for this way of working.

01

Run from anywhere

Start a one-shot task or an ongoing chat directly in the shell you already use.

02

Find local models

Discover Ollama, LM Studio, llama.cpp, and compatible endpoints before choosing a profile.

03

Automate safely

Use workspace commands, plans, memory, MCP configuration, and the JSON-lines service without coupling your workflow to a UI.

Get started

Install with npm

Requires Node.js 20 or newer. The same workspace profiles work across the CLI and TUI.

$ npm install -g @truss-harness/cli
$ truss-cli models
$ truss-cli config init
$ truss-cli chat "Review the current diff"
View CLI on npmComplete documentation

A typical flow

  1. 01

    Discover a server and model with truss-cli models.

  2. 02

    Create a workspace profile with truss-cli config init.

  3. 03

    Run a task, inspect streamed tool activity, and keep the workspace state for the next session.