Get Started

Requirements

• Node.js 20 or newer
• npm 10 or newer
• A local model server such as Ollama, LM Studio, or llama.cpp
• Git for commit-message generation

Install

npm install -g @truss-harness/cli @truss-harness/tui

The CLI is enough to begin. Install the terminal UI when you want the full-screen workspace client.

Confirm the executables are on your path:

truss-cli help
truss-tui

The npm global install prints a short CLI checklist. Running truss-cli with no arguments always displays the full command reference.

Interactive setup

Run truss-cli setup to detect local servers, choose a model, and save a reusable user-level default profile. The wizard is a command instead of an install-time prompt so npm installs remain safe in CI and scripts.

Find a local model

truss-cli models

When no model has been configured, Truss probes standard local endpoints and selects a usable model automatically. An actively running Ollama model has priority.

Run a first chat

truss-cli chat "Explain this workspace."

If no local server is running, configure a profile first. See Configuration.

Run a first agent task

Run Agent/Edit mode only inside a workspace you trust. Direct CLI chat is non-interactive and auto-allows tools.

truss-cli chat --mode edit "Read README.md and add one concise sentence describing the CLI. Make no other changes."

Use the TUI, desktop app, or VS Code extension when you want interactive approval prompts and persistent conversation controls.

Continue setup

Create a reusable local-model profile with truss-cli config init, continue to Local Models, or see Troubleshooting when discovery or tool calling fails.