Getting Started

This guide helps you install GPTCode, configure providers, and start using the core workflows. It also includes a 10‑second quick start for universal feedback capture.

Install

Initial setup

gt setup             # creates ~/.gptcode/setup.yaml
gptcode key openrouter    # add API key(s) as needed
gt backend           # check current backend
gt backend list      # list all backends
gt profile           # check current profile

Quick start: two‑keystroke feedback (Ctrl+g)

Capture corrections from any CLI as training signals.

# zsh
gt feedback hook install --with-diff --and-source

# bash
gt feedback hook install --shell=bash --with-diff --and-source

# fish
gt feedback hook install --shell=fish --with-diff

Usage: 1) Type/paste the suggested command 2) Press Ctrl+g to mark the suggestion 3) Edit (if needed) and press Enter

GPTCode records good/bad outcomes and saves changed files and optional git patch.

Check stats:

gt feedback stats

Core commands

Important: Not all models support tool calling. Use these recommended models for best results:

Backend Model Context Best For
Groq groq/compound 128k General use, tool calling ✅
Groq llama-3.3-70b-versatile 128k Fast, large context
OpenRouter openrouter/auto varies Auto-routing, best model per task
Ollama qwen2.5-coder:32b 32k Local, tool calling ✅
Ollama llama3.3:70b 128k Local, large context
# Set your default model
gt model set groq/compound

# Or for local
gt model set qwen2.5-coder:32b

Troubleshooting