Skip to main content
AgentUse Logo

​
AI Agents on Autopilot

AgentUse is a CLI framework for building AI agents that run themselves. Define agents in markdown, run via cron, CI/CD, or serverless, and get results without babysitting. No SDK. No flowcharts. Just AI that works while you don’t.

Quick Start

Get up and running with your first agent in under 5 minutes

Examples

Browse ready-to-use agent examples

​
Why AgentUse?

​
The Problem

AI tools today force you to choose:
  • Interactive copilots require your constant attention - great for pair programming, but can’t run unattended
  • Visual workflow tools give you drag-and-drop simplicity but create version control nightmares and vendor lock-in
  • SDK-heavy frameworks offer power but require hundreds of lines of boilerplate just to say “hello world”

​
The Solution

AgentUse puts AI agents on autopilot. Define agents in markdown, run via cron, CI/CD, or serverless, and get results asynchronously.
Run via cron jobs, CI/CD pipelines, and serverless functions. No babysitting required.
Diff, review, and merge agents like any other code. Your agents are text files that integrate seamlessly with Git.
If you can write plain English, you can build an agent. No new syntax or programming required.
Built-in retries, streaming, error recovery, and MCP support. Sub-second startup times.
Copilots assist you. AgentUse agents work for you.

​
Key Features

1

Autopilot Execution

Run agents via cron jobs, CI/CD pipelines, and serverless functions. Sub-second startup times.
2

Multi-Provider Support

Works with Anthropic Claude (including OAuth), OpenAI GPT, and OpenRouter with flexible API key management.
3

MCP Integration

Connect to any Model Context Protocol server for database, filesystem, and API access.
4

Sub-Agent Support

Build complex workflows by composing multiple specialized agents.

​
How It Works

AgentUse agents are markdown files with YAML configuration and plain English instructions: 
---
model: anthropic:claude-haiku-4-5
---

You are a helpful assistant.

## Task
Say hello to the user and ask how you can help them today.
That’s it! Save this as hello.agentuse and run: 
npx -y agentuse@latest run hello.agentuse
The agent name is automatically derived from the filename, keeping configuration minimal.

​
What Can You Build?

  • MCP-Powered Tools - Access databases, filesystems, and APIs through MCP servers
  • Multi-Agent Systems - Compose specialized agents for complex workflows
  • Automation Scripts - Replace complex bash scripts with intelligent agents
  • Data Processing - Transform and analyze data with natural language
  • Content Generation - Create content with consistent voice and style
  • Integration Workflows - Connect multiple services through MCP protocols