python-sdk/docs/getting-started/quickstart.md

Quickstart

Get up and running with Hanzo Python SDK in minutes.

1. Install

pip install hanzo-mcp

2. Run MCP Server

With Claude Code

uvx hanzo-mcp

Claude Code will auto-detect and use all 30+ tools.

With VS Code Extension

1. Install the Hanzo extension
2. The extension auto-detects uvx and starts the Python MCP
3. All tools are available in the AI assistant

3. Use the Tools

File Operations

# Read a file
read(file_path="/path/to/file.py")

# Edit a file
edit(
    file_path="/path/to/file.py",
    old_string="old code",
    new_string="new code"
)

# Search for patterns
search(pattern="TODO", path="./src")

Shell Commands

# Run commands (auto-backgrounds after 30s)
cmd("npm install")

# Run in parallel
cmd(["npm install", "cargo build"], parallel=True)

# DAG execution
cmd([
    "mkdir dist",
    {"parallel": ["cp a dist/", "cp b dist/"]},
    "zip -r out.zip dist/"
])

Browser Automation

# Navigate to page
browser(action="navigate", url="https://example.com")

# Click element
browser(action="click", selector="button.submit")

# Take screenshot
browser(action="screenshot", full_page=True)

# Mobile emulation
browser(action="emulate", device="mobile")

Memory & Reasoning

# Save to memory
memory(action="create", data={"note": "Important insight"})

# Recall memories
memory(action="recall", query="project architecture")

# Structured thinking
think(thought="Analyzing the problem...")

# Critical analysis
critic(analysis="Review this implementation...")

4. Agent SDK

Build your own AI agents:

from agents import Agent, Runner

# Create an agent
agent = Agent(
    name="code_reviewer",
    instructions="""
    You are a code review expert.
    Analyze code for bugs, performance issues, and best practices.
    """,
    tools=[review_code, suggest_improvements]
)

# Run the agent
result = Runner.run_sync(
    agent,
    "Review this Python function for issues..."
)

print(result.final_output)

Multi-Agent Systems

from agents import Agent, handoff

# Create specialized agents
security_agent = Agent(
    name="security",
    instructions="Analyze code for security vulnerabilities."
)

performance_agent = Agent(
    name="performance",
    instructions="Analyze code for performance issues."
)

# Main coordinator
lead_agent = Agent(
    name="lead",
    instructions="Coordinate code review. Handoff to specialists.",
    handoffs=[
        handoff(security_agent, "security issues"),
        handoff(performance_agent, "performance concerns")
    ]
)

5. Configuration

Environment Variables

# Disable auto-backgrounding
export HANZO_AUTO_BACKGROUND_TIMEOUT=0

# Set allowed paths
export HANZO_ALLOWED_PATHS="/home/user/projects,/tmp"

VS Code Settings

{
  "hanzo.mcp.backend": "python",
  "hanzo.mcp.pythonCommand": "uvx hanzo-mcp",
  "hanzo.mcp.disableBrowserTool": false,
  "hanzo.mcp.enabledTools": ["read", "write", "cmd", "search"]
}

Next Steps

• MCP Tools Reference - Complete tool documentation
• Agent SDK Guide - Build custom AI agents
• Configuration - Advanced configuration options