Choose a solution by outcome
You do not need to learn Zhin's package graph first. Pick the outcome you need; each path starts with a runnable minimum and adds capabilities only when they are useful.
| What you want to deliver | Start here | You are done when |
|---|---|---|
| A bot that receives and sends messages | IM Bot path | Sandbox and a real platform share the same command and message flow |
| A bot that understands your business | Add governed Agent context | The Prompt Section is in the current generation and visible in Console |
| Tools, MCP, skills, and sub-agents | AI Agent path | The Agent sees only capabilities authorized for the current turn |
| Operate and diagnose a deployed bot | Console Admin path | You can inspect endpoints, logs, capabilities, and Agent runs |
| Operate one product through multiple accounts | Multi-Endpoint operations | Each account is diagnosable while business capabilities stay shared |
| Give an Agent business context without excess authority | Governed business Agent | Prompts, tools, approvals, and execution policy are independently auditable |
| Use one GitHub repository as a collaboration space | GitHub repository Workroom | Repository events enter the right Project and task projection boundaries are explicit |
| A personal schedule, reminder, and voice assistant | Personal Assistant | Messages, scheduled jobs, voice, and proactive delivery form one loop |
| A multi-platform community bot | Multi-platform Bot | Business commands and components do not depend on one adapter |
Recommended delivery order
- Establish a repeatable golden path in Sandbox.
- Connect one real platform without leaking adapter details into business code.
- Enable AI only when needed; govern prompts, tools, and authority separately.
- Use Console to verify the generation that is actually published instead of treating the configuration file as runtime truth.
If you do not have a project yet, use the default IM golden path in Quick Start.