​
Zaby SDK

The official SDK for the Zaby AI Agent Platform. Use Python or TypeScript to create agents, manage deployments, stream runtime responses, and integrate with external tools.
Select your preferred language tab for code examples: Python or TypeScript.

Quick Start

Install, authenticate, and run your first agent in under 5 minutes.

Configuration

Timeouts, retries, dynamic auth providers, and environment setup.

Client Reference

Full API reference for all clients — agents, runtime, MCP, memory, and more.

Error Handling

Error types, status codes, retry strategies, and best practices.

​
Installation

pip install zaby-sdk
Requires Python 3.10+ or Node 20+.

​
Minimal Example

import asyncio
from zaby import Zaby

async def main():
    zaby = Zaby(api_key="zaby_pk_...")

    agent = await zaby.agents.create({
        "name": "my-agent",
        "slug": "my-agent",
    })
    print(f"Agent created: {agent['id']}")

    health = await zaby.health.check()
    print(f"Server status: {health['status']}")

asyncio.run(main())

​
Two Client Modes

The SDK provides two client classes for different contexts:

Zaby

Admin-level operations: create agents, manage deployments, configure MCP servers, create runtime tokens.

Auth: API key (x-zaby-api-key) + optional JWT

ZabyRuntime

Runtime agent interactions: start runs, stream SSE events, submit feedback and approvals.

Auth: Runtime token JWT (authorization: Bearer)

​
What’s Next?