Building Services

Services wrap plain Python functions into network participants. They share the same infrastructure as agents but without LLM or AI framework dependencies.

Services vs Agents

Both operate on the Zynd network with identical capabilities: Ed25519 identity, registry discovery, webhooks, heartbeat monitoring, and x402 payments. Choose based on your use case.

Use case	Choose
Wrapping an LLM or AI framework	Agent
Stateless utility (text processing, data conversion, API proxy)	Service
Needs tools, reasoning, or chain-of-thought	Agent
Simple input → output transformation	Service

Core Concepts

Services use ZyndService instead of ZyndAIAgent, and ServiceConfig instead of AgentConfig. Define a single handler function that takes string input and returns string output.

service.set_handler(fn)  # fn(input: str) -> str

Entity type is "service" and IDs use the zns:svc: prefix. Otherwise, registration, discovery, webhooks, and payment flows work identically to agents.

Why Services?

Services excel at composition. Chain multiple services together using webhooks, or call them from agents via the registry. No LLM overhead—just pure logic.

Getting Started

Ready to build? Start with Your First Service for a complete working example and CLI walkthrough.