1
What it is
它是什么
Goku-AIOS is a self-hosted enterprise AI agent platform split into two independently deployable services that share one database. Goku-Core is the user-facing runtime — conversations, tasks, approvals, and omnichannel delivery. Goku-Studio is the developer control plane — agent authoring, workflow design, tools, MCP servers, and knowledge management.
Goku-AIOS 是面向企业的私有化 AI Agent 平台，已拆分为两个独立部署、共享数据库的服务。Goku-Core 是面向用户的运行时 — 对话、任务、审批和全渠道交付。Goku-Studio 是研发控制台 — Agent 配置、工作流设计、工具、MCP 服务器和知识库管理。
-
Build agents, workflows, and tools in Studio without touching the production runtime.
在 Studio 中构建 Agent、工作流和工具，无需接触生产运行时。
-
Private deployment for regulated or security-sensitive environments.
适合受监管或对安全敏感环境的私有化部署。
-
JWT single sign-on — log in once through Core, access both planes.
JWT 单点登录 — 通过 Core 登录一次，两个平面均可访问。
2
Why it is different
为什么它不同
Most platforms bundle authoring and runtime into one service. Goku-AIOS separates them so developer iteration in Studio never risks production stability in Core — while both stay in sync through a shared database.
大多数平台把编写和运行时捆绑在一个服务中。Goku-AIOS 将二者分离，Studio 中的研发迭代不会影响 Core 的生产稳定性，同时通过共享数据库保持同步。
-
Self-evolving agents driven by reactions, outcomes, and A/B improvements.
基于反馈、结果和 A/B 改进机制驱动的自我进化 Agent。
-
Omnichannel messaging across Feishu, Teams, Email, and mobile web.
覆盖飞书、Teams、邮件和移动 Web 的全渠道消息层。
-
Approval chains, audit logs, RBAC, and DLP built into Core; agent configuration owned by Studio.
Core 内置审批链、审计日志、RBAC 和 DLP；Agent 配置由 Studio 统一管理。
3
Where it fits
它适合哪里
The strongest fit is for enterprises that want a clean separation between who builds agents (Studio) and who uses them (Core) — with governance, channels, and audit all built in to the runtime side.
它最适合那些希望明确区分"谁构建 Agent"（Studio）和"谁使用 Agent"（Core）的企业，同时在运行时侧内置治理、渠道和审计能力。
-
Dev teams iterating on agents in Studio while ops teams run them in Core.
研发团队在 Studio 迭代 Agent，运营团队通过 Core 运行。
-
Customer support and operations running across multiple channels.
跨多渠道运行的客服与运营场景。
-
Executive reporting, recurring analysis, and monitored workflows.
管理层报告、周期性分析和带监控的工作流。
4
How to evaluate it
如何评估它
Start both services with Docker Compose (Core on :5106, Studio on :5107). Log in through Core, open Studio from the sidebar, and build your first agent. The runtime picks it up immediately from the shared database.
通过 Docker Compose 同时启动两个服务（Core 监听 :5106，Studio 监听 :5107）。通过 Core 登录后，从侧边栏进入 Studio，构建第一个 Agent。运行时会立即从共享数据库中读取配置。
-
Docker Compose starts Core + Studio + MySQL in one command.
Docker Compose 一条命令同时启动 Core、Studio 和 MySQL。
-
Move to Helm and Kubernetes for scaled, independent scaling of each service.
迁移到 Helm 和 Kubernetes 后，可对每个服务独立扩容。
-
Use the Studio API, SDK, and docs for technical due diligence.
结合 Studio API、SDK 和文档完成技术尽调。
