Getting started

Introduction

DirectOS is an AI operations platform: build AI workflows visually, execute them deterministically, and optimize them continuously.


What is DirectOS?

DirectOS is an AI-native workflow automation platform and part of the Bezpośrednik ecosystem — automation without middlemen. You compose workflows on a visual canvas, a deterministic engine executes them with retries and parallelism, and a multi-LLM router sends every AI task to the cheapest model that clears the quality bar.

The part that makes it an operations platform is what happens after the first run: cost attribution per call, provider health, routing explanations and concrete recommendations — the feedback loop we call execution engineering. Building a workflow is a day of work; running it efficiently for a year is the actual problem.

The platform at a glance

ComponentWhat it is
DirectFlowWorkflow engine — parallel scheduling, retries, branching, expressions
DirectRouterMulti-LLM router — ten providers, cost-aware, ordered fallbacks
DirectStudioVisual editor, execution history and the AI Operations Center
DirectAgentsTool-calling agents with reflection, delegation and messaging
DirectMemory & RAGSemantic memory and document knowledge bases with hybrid search
LLM GatewayAnthropic-compatible endpoint with routing, budgets and call logs
Built-in nodesTriggers, logic, data, AI and integrations — 25 in total

OAuth / single sign-on, multi-tenant organizations and a node marketplace are still on the roadmap — nothing in the product fakes them. See the public roadmap.

How it is layered

architecture
┌───────────────────────────────────────────────────────────────┐
│ DirectStudio — visual editor + AI Operations Center           │
│   REST  ·  live events over WebSocket                         │
├───────────────────────────────────────────────────────────────┤
│ DirectOS API — workflows │ executions │ agents │ webhooks     │
│                memory │ RAG │ credentials │ auth │ gateway    │
├───────────────────────────────────────────────────────────────┤
│ Built-in nodes          │ DirectRouter (providers, cost)      │
├───────────────────────────────────────────────────────────────┤
│ DirectFlow — workflow engine (pure, deterministic)            │
└───────────────────────────────────────────────────────────────┘

The dependency rule: arrows only point downward. The engine knows nothing about HTTP or databases, which is what keeps runs deterministic and testable; the API composes engine, nodes and router; DirectStudio uses the same public API your scripts do.

Where to go next

  • Quickstart — the demo, your first run and your first API call.
  • Workflows — how the engine schedules, branches and retries.
  • Model guide — the full catalog and the cost-optimization playbook.
  • API reference — every endpoint, plus live execution events.