Introduction

What is UnderOcean?

UnderOcean is an enterprise AI orchestration platform for building, running, and governing AI agents, automated workflows, and knowledge-grounded (RAG) assistants — all within a single, secure, multi-tenant workspace built for organizational use.

Getting Started

Set up your workspace and run your first agent.

Concepts

Agents, flows, providers, RAG, MCP, context management.

Guides

Step-by-step walkthroughs for common tasks.

API Reference

REST endpoints, streaming events and error formats.

Platform capabilities

Capability	Description
Multi-tenant workspaces	Organizations, members, and projects with strict data isolation between tenants
Agent Builder	Configure an agent's behavior, model, and tool access — no code required
Flow Builder	Design multi-step automations visually on an easy-to-use canvas
Knowledge Base (RAG)	Ground agent answers in your own documents with per-project knowledge bases
Tool & Integration Layer	Connect external tools and services through a standardized integration protocol
Bring your own model	Use any major AI provider, or point to your own private/self-hosted model
Context Management	Agents automatically retain relevant history and long-term memory
Real-time Streaming	Responses and reasoning steps stream live as agents work
Role-Based Access Control	Fine-grained permissions across organizations, projects, and custom roles

How UnderOcean is built

UnderOcean runs as a fully self-contained platform: an orchestration engine that executes agents and workflows, a secure data layer that enforces tenant isolation at every level, a pluggable model layer that lets you connect any AI provider, and a vector-based knowledge layer for retrieval-augmented generation. Every capability is designed to be governed, audited, and scaled for organizational use.

See Architecture → System Overview for a closer look at how these pieces fit together, including tenancy and the security model.