JSPM

  • ESM via JSPM
  • ES Module Entrypoint
  • Export Map
  • Keywords
  • License
  • Repository URL
  • TypeScript Types
  • README
  • Created
  • Published
  • Downloads 5
  • Score
    100M100P100Q102174F
  • License MIT

Usda Code - Open-source AI coding agent with terminal-first experience and multi-model support

Package Exports

  • usda-code
  • usda-code/dist/cli/index.js

This package does not declare an exports field, so the exports above have been automatically detected and optimized by JSPM instead. If any package subpath is missing, it is recommended to post an issue to the original package (usda-code) to support the "exports" field. If that is not possible, create a JSPM override to customize the exports field for this package.

Readme

Usda Code

Open-source AI coding agent with a terminal-first experience, multi-model support, and full tool execution.

Write code, run commands, manage files — all from your terminal with any LLM provider.

✨ Features

  • Terminal-First UI — Full-screen interactive interface with animated feedback, keyboard shortcuts, and word-by-word response streaming
  • Multi-Model Support — Works with OpenAI, Anthropic, Google, NVIDIA, OpenRouter, Ollama, vLLM, or any OpenAI-compatible API
  • Dual Agent Modes — Full execution mode for active coding, planning mode for safe analysis
  • Tool System — Filesystem operations, shell commands, and git integration with structured validation
  • Session Persistence — Resume conversations across restarts with usda -r <sessionId>
  • Plugin Architecture — Extend with custom tools and providers
  • Human-in-the-Loop — Approval system for destructive operations; toggle with -a for auto-approve
  • Cross-Platform — macOS, Windows, Linux

📦 Installation

npm i -g usda-code

Requires Node.js 18+.

🚀 Quick Start

# Interactive mode — opens the full terminal UI
usda

# Run a single task directly
usda "create a REST API with Express"

# Auto-approve all tool actions (no y/n prompts)
usda -a

# Resume a previous session
usda -r <sessionId>

On first run, usda will walk you through selecting a provider, model, and API key. Configuration is saved to usda.config.yaml in the current directory.

⌨️ Keyboard Shortcuts

Key Action
Tab Autocomplete /commands
Ctrl+O Expand/collapse latest thinking
Ctrl+E Expand/collapse latest prompt
Ctrl+J Toggle all collapsed/expanded items
Ctrl+C ×2 Exit (double-press within 2s)

🔧 Configuration

Create usda.config.yaml in your project root (or let usda generate it on first run):

provider:
  provider: openrouter
  model: anthropic/claude-sonnet-4-20250514
  apiKey: sk-or-...
  baseUrl: https://openrouter.ai/api/v1
  temperature: 0.7
  maxTokens: 4096

agent:
  mode: full            # full | planning
  maxSteps: 50
  requireApproval: false
  blockedCommands:
    - rm -rf
    - format
    - shutdown
  commandTimeout: 30000

🔌 Supported Providers

Provider Base URL Example Models
OpenAI https://api.openai.com/v1 gpt-4o, gpt-4o-mini
Anthropic https://api.anthropic.com claude-sonnet-4-20250514, claude-3-haiku
Google https://generativelanguage.googleapis.com/v1beta/openai gemini-2.0-flash
NVIDIA https://integrate.api.nvidia.com/v1 meta/llama3-70b-instruct
OpenRouter https://openrouter.ai/api/v1 Any model on OpenRouter
Ollama http://localhost:11434/v1 llama3.1, mistral, deepseek-coder
vLLM http://localhost:8000/v1 Any locally hosted model
Custom User-defined Any OpenAI-compatible endpoint

📝 Slash Commands

Type these during interactive mode:

Command Description
/help Show all commands
/model Change the current model
/provider Switch provider
/key Update API key
/mode Toggle between full and planning mode
/auto Toggle auto-approve
/status Show session stats (steps, tokens)
/clear Clear the screen
/new Start a new session
/q Quit

🛠️ What Usda Code Can Do

  • Read, write, create, and edit files in your project
  • Run shell commands — install deps, run builds, execute tests
  • Search and refactor codebases — grep, analyze, restructure
  • Git operations — stage, commit, diff, log
  • Debug issues — trace errors, inspect logs, fix code
  • Multi-step tasks — break down complex work into logical steps

🏗️ Architecture

src/
├── agents/        # Full & Planning agent implementations
├── approval/      # Human-in-the-loop approval system
├── cli/           # Terminal UI (renderer, input handling)
├── config/        # YAML config loader & types
├── plugins/       # Plugin system (extensible tools)
├── providers/     # LLM provider adapters (OpenAI, Anthropic, etc.)
├── session/       # Session persistence & memory
├── tools/         # Built-in tools (filesystem, shell, git)
└── utils/         # Logger, helpers, env

📄 License

MIT