agent-dev-cli 0.0.1b260303

AI Toolkit - CLI and SDK for agent debugging and workflow visualization

Agent Dev CLI

A Python CLI and SDK for agent debugging and workflow visualization with VS Code AI Toolkit integration.

Installation

pip install agent-dev-cli --pre

Note: This package is currently in beta. The --pre flag is required to install pre-release versions.

Usage

Option 1: CLI Wrapper (Recommended)

The easiest way to use Agent Dev CLI - no code changes required:

# Run your agent script with agentdev instrumentation
agentdev run my_agent.py

# Specify a custom port
agentdev run workflow.py --port 9000

# Enable verbose output
agentdev run my_agent.py --verbose

# Pass arguments to your script
agentdev run my_agent.py -- --model gpt-4 --temperature 0.7

The CLI automatically:

• Intercepts from_agent_framework() calls
• Injects agentdev visualization endpoints
• Opens the workflow visualization in VS Code

Option 2: Programmatic API

If you prefer explicit control, you can integrate agentdev directly:

from agentdev import configure_agent_server
from azure.ai.agentserver.agentframework import from_agent_framework

# Create your agent
agent = build_agent(chat_client)

# Create agent server
agent_server = from_agent_framework(agent)

# Setup workflow visualization
configure_agent_server(agent_server)

# Run the server
await agent_server.run_async()

CLI Commands

agentdev run

Run a Python agent script with agentdev instrumentation.

agentdev run [OPTIONS] SCRIPT [ARGS]...

Options:
  -p, --port INTEGER    Agent server port (default: 8087)
  -v, --verbose         Enable verbose output
  --help                Show this message and exit

agentdev info

Show agentdev configuration and status information.

agentdev info

Features

• Health Check Endpoint: Adds a /agentdev/health endpoint to your agent server
• Workflow Visualization: Starts a visualization server on port 8090 for WorkflowAgent instances
• Easy Integration: Simple one-function setup

Testing Your Agent Server

Once your agent server is running, you can test it using curl:

# Health check
curl http://localhost:8087/agentdev/health

# Send a request to your agent (streaming)
curl 'http://localhost:8087/agentdev/v1/responses' \
  -H 'Content-Type: application/json' \
  -d '{"model":"your-agent-model-id","input":{"role":"user","text":"Hello!"},"stream":true}'

Requirements

• Python 3.10+
• VS Code AI Toolkit (for visualization)

License

This project is licensed under the Microsoft AI Toolkit – Agent Dev CLI License Terms.