Skip to main content
Rowboat CLI (rowboatx) is a command-line tool for building, running, and managing AI agents and workflows. It provides a local-first approach to AI automation with support for multiple LLM providers and Model Context Protocol (MCP) servers.

Key Features

  • Agent Runtime: Run AI agents locally with interactive or non-interactive modes
  • Workflow Management: Import, export, and share workflow configurations
  • MCP Integration: Connect to Model Context Protocol servers for extended capabilities
  • Multi-Provider Support: Works with OpenAI, Anthropic, Google, Ollama, and more
  • Interactive UI: Optional terminal-based dashboard for managing runs
  • Local Configuration: All data stored in ~/.rowboat directory

Quick Start

1

Install the CLI

Install rowboatx globally via npm:
npm install -g @rowboatlabs/rowboatx
2

Configure your model

Set up your preferred LLM provider:
rowboatx model-config
This interactive wizard will guide you through selecting a provider (OpenAI, Anthropic, Google, etc.) and configuring API keys.
3

Run your first agent

Start the default copilot agent:
rowboatx
Or specify a custom agent:
rowboatx --agent my-agent

Architecture

Configuration Directory

All Rowboat CLI data is stored in ~/.rowboat/:
~/.rowboat/
├── agents/          # Agent definitions (JSON)
├── config/
│   ├── models.json  # LLM provider configuration
│   └── mcp.json     # MCP server definitions
└── runs/            # Run logs (JSONL)

Core Components

Agents

JSON-based agent definitions with instructions, tools, and workflows

MCP Servers

External tool providers following the Model Context Protocol

Run Engine

Execute agents with state management and event streaming

API Server

Optional HTTP server for programmatic access

Use Cases

  • Development Automation: Create agents that help with coding tasks
  • Data Processing: Build workflows for ETL and data transformation
  • Content Generation: Automate writing and content creation tasks
  • Integration Testing: Run agents in CI/CD pipelines with --no-interactive mode
  • Personal Assistant: Local AI assistant with access to your tools and data

Next Steps

Installation

Detailed installation and setup instructions

Commands

Complete command reference

Build docs developers (and LLMs) love