acai

Overview

Acai is an AI-driven command-line tool that assists software developers with coding, debugging, refactoring, and workflow automation. It provides both a conversational REPL and a modern TUI for interacting with large language models in the context of your codebase.

Key Capabilities

• Interactive AI Assistant: Conversational REPL and TUI with modal dialogs, autocomplete, and rich text formatting.
• Codebase Interaction: Read, edit, search, and navigate files with context-aware AI assistance.
• Git Integration: Generate conventional commits, review pull requests, and manage local changes.
• Multi-Model Support: Switch between OpenAI, Anthropic, Google, DeepSeek, Groq, X.AI, OpenRouter, OpenCode Zen, and OpenCode Go.
• Piped Input: Pipe text via stdin for REPL mode or as context with the -p flag.
• Skills System: Discover and load specialized instruction files for specific tasks.
• Multi-workspace Support: Work across multiple project directories simultaneously.

Prerequisites

• Node.js 24 or higher
• Git
• Ripgrep (rg) - Fast file content searching
• GitHub CLI (gh) - Git operations and repository management

# macOS
brew install ripgrep gh

# Ubuntu/Debian
sudo apt install ripgrep gh

Installation

npm install -g @travisennis/acai

Quick Start

# Start interactive mode
acai

# Specify a model
acai --model anthropic:sonnet

# One-shot CLI mode
acai -p "What files contain the term 'toolCallRepair'?"

# Pipe input
echo "How many TypeScript files are in this project?" | acai

# Resume a previous session
acai --resume

Once in the REPL, type prompts or use commands:

> How do I read a file in Node.js?
> @source/index.ts
> /help

Reference files directly with @filename, directories with @dirname, or run shell commands with !`command`.

Technologies

• TypeScript and Node.js
• AI SDK (@ai-sdk/*) for LLM provider integration
• Ripgrep for fast file content searching
• Pino for structured logging
• Zod for schema validation
• Biome for formatting and linting
• Fallow for dead code, duplication, and complexity analysis

Project Structure

.
├── source/            # Main application source code
│   ├── agent/         # Agent loop
│   ├── cli.ts         # CLI entry point
│   ├── commands/      # REPL command implementations
│   ├── execution/     # Command execution utilities
│   ├── middleware/     # AI request/response middleware
│   ├── models/        # AI model providers and management
│   ├── prompts/       # Prompt generation and management
│   ├── repl/          # REPL utilities
│   ├── sessions/      # Session persistence and management
│   ├── terminal/      # Terminal output formatting and rendering
│   ├── tui/           # Terminal User Interface components
│   ├── tools/         # AI-callable tools (filesystem, git, web, bash, etc.)
│   ├── tokens/        # Token counting and tracking
│   └── utils/         # Utility functions
├── test/              # Unit tests
├── docs/              # Additional documentation
├── ARCHITECTURE.md    # Detailed architectural overview and flow diagrams
├── CONTRIBUTING.md    # Development setup and guidelines
└── AGENTS.md          # Project-specific AI rules and guidelines

Documentation

• Usage Guide - Commands, keyboard shortcuts, piped input, and prompt syntax
• Configuration - Environment variables, project and global settings
• Skills System - Creating and using specialized instruction files
• Dynamic Tools - Creating custom tools to extend acai (supports bash, python, and other languages; Amp-compatible text schema format; /tools make scaffolding command)
• Architecture - Internal architecture and flow diagrams
• Contributing - Development setup, scripts, and code style
• Agent Instructions - Routing guide for AI agents working in this repo
• Agent Guardrails - Focused compatibility, security, testing, and documentation rules

Acknowledgments

Special thanks to OpenCode for providing access to their AI model aggregation service, which enables seamless integration with multiple state-of-the-art models through a unified API.

License

This project is licensed under the MIT License - see the LICENSE file for details.