forge-coder 0.2.2

Local-first autonomous coding agent CLI

forge — Local Coding Agent

Local-first autonomous coding agent CLI.

Documentation

• Prerequisites — Before you start (LM Studio, API keys)
• User Manual — Command reference and workflows

Install

One line (Linux/macOS)

curl -fsSL https://raw.githubusercontent.com/Tushaarxr/forge/main/install.sh | bash

One line (Windows PowerShell)

irm https://raw.githubusercontent.com/Tushaarxr/forge/main/install.ps1 | iex

Or with pipx

pipx install forge-coder

Or with Docker

docker build -t forge-agent .
docker run -it -e GEMINI_API_KEY=AIza... -v $(pwd):/workspace forge-agent auto "build a todo app"

# With docker-compose
GEMINI_API_KEY=AIza... docker-compose run forge auto "build a todo app"

Quickstart (3 steps)

forge setup         # one-time wizard: API keys + LM Studio check
cd my-project/
forge auto "build a FastAPI todo app with SQLite and JWT auth"

Commands

Command	Description
forge setup	One-time setup: Configure API keys and create environment files.
forge init	Initialize forge: index files into vector store and build dependency graph.
forge run	Run the forge coding agent on a goal.
forge auto	Autonomous end-to-end build mode.
forge chat	Interactive REPL: chat with the coding agent.
forge status	Show current project status and metrics.
forge summarise	Create a checkpoint summary of recent changes.
forge rollback	Rollback files to their .forge_backup versions.

Requirements

• Python 3.10+
• LM Studio (free): https://lmstudio.ai — runs Qwen3.5-9B locally
• Gemini API key (free): https://aistudio.google.com

Architecture

Forge leverages a "Master Brain" (Gemini) for high-level project planning and a "Local Worker" (LM Studio / Qwen3.5) for fast, iterative code execution. It uses FAISS for vector search and NetworkX to map your project's dependency graph.

Contributing / Development setup

git clone https://github.com/Tushaarxr/forge.git
cd forge
pip install -e ".[dev]"
pytest tests/