An AI development agent built with agentlys.

Give it a task and it will autonomously code, test, and deploy using real development tools.

Think about it as a simple Claude Code but for any LLM provider.

• Code Editor: Read, write, edit files with syntax awareness
• Terminal: Run commands, scripts, build tools, tests
• Git Operations: Branch, commit, push, create PRs
• Web Rendering: Take screenshots of web apps for testing
• Multi-Agent: Coordinate multiple specialized agents

pip install -e .

Requires:

• Python 3.9+
• OpenAI or Anthropic API key

# Start with a task
agentlys-dev "Create a FastAPI app with user authentication"

# Or start interactive mode
agentlys-dev
> Create a React component for file upload
> Add unit tests
> Deploy to Vercel

The project exposes a few other convenience commands:

agentlys-dev                    # Interactive agent
agentlys-dev-dual              # Dual-agent collaboration demo
agentlys-dev-export            # Export conversation history
agentlys-dev-github-issue-server  # GitHub webhook server

Built on agentlys, agentlys-dev gives AI agents access to real development tools:

from agentlys import Agentlys
from agentlys_tools.code_editor import CodeEditor
from agentlys_tools.terminal import Terminal

agent = Agentlys(
    instruction="You are a senior developer...",
    provider="anthropic",
    model="claude-sonnet-4-20250514"
)

# Add development tools as classes
agent.add_tool(CodeEditor(), "CodeEditor")
agent.add_tool(Terminal(), "Terminal")

# Agent now has full development capabilities
for message in agent.run_conversation("Build a web scraper with tests"):
    print(message.to_markdown())

Choose your provider and set your api key:

export ANTHROPIC_API_KEY="your-key"

export OPENAI_API_KEY="your-key"