Core Concepts

qRaptor organizes your work into a clear hierarchy. Understanding these concepts helps you navigate the Studio confidently.

Workspace

A workspace is your top-level account — the billing and isolation boundary. Every resource you create belongs to exactly one workspace, and your data is completely isolated — never visible to other workspaces.

Each workspace is on a plan (Free, App Lite, Starter, Pro, Team, or Enterprise) that determines features, credits, and team size.

Project

A project groups related agents, apps, data, tools, and deployments. Think of it as a container for a single product or initiative. You can have multiple projects per workspace (the number depends on your plan).

Inside a project, you find:

Agents

An agent is an AI-powered execution unit. You build agents visually on a drag-and-drop canvas. qRaptor supports two agent types:

Task Agent

A Task Agent executes a defined workflow as a graph of steps. You connect nodes (start → tools → logic → output) to build deterministic pipelines. Ideal for structured processes: data extraction, form processing, scheduled jobs, and multi-step automations.

AI Agent

An AI Agent is conversational and autonomous. It reasons about requests, uses tools, accesses memory, and responds in natural language. You configure its persona, tool access, and knowledge — then it handles conversations intelligently. Ideal for customer support, sales assistants, and interactive copilots.

Multi-Agent Systems (MAS)

A MAS coordinates multiple agents working together. The orchestration pattern is the Conversation Router: an LLM classifies each incoming request against the member agents’ skills and routes it to the best specialist, carrying the conversation across handoffs so it stays seamless. See Multi-Agent System overview.

Apps (Vibe Code)

Apps are full-stack applications generated from natural language descriptions. The platform pipeline works in visible stages:

Prompt → Requirements → App Graph → Code → Auto-Fix → Iterate

You can generate:

• Web apps — React full-stack applications
• Mobile apps — React Native / Expo applications
• Copilots — Embeddable AI chat widgets powered by your agents

Each app has its own workspace with code editor, live preview, build status, and App Context Graph.

Data Vault

The Data Vault is your secure data layer:

• Tables — Structured data with schema design, ER diagrams, and a query explorer.
• Documents — File storage for uploads, exports, and attachments.
• Content Stores — Vector stores for RAG (retrieval-augmented generation). Your agents search these for relevant knowledge.
• Memory — Three tiers of agent memory:
  • Profile — Remembers user preferences across conversations.
  • Episodic — Recalls what happened in past sessions.
  • Semantic — Self-learned knowledge the agent discovers over time.

Nodes

A node is a single step in a Task Agent’s graph. qRaptor provides node types across these categories:

• Flow Control — Start, End
• Logic — Condition, Switch
• Loops — Loop
• Parallel — Parallel, Merge
• Actions — Quick Prompt (call an LLM), Tools Call, Agent Call (invoke another agent)
• Data — Query Table, Insert Row, Update Row, Delete Row
• Memory — Memory Read, Memory Write
• RAG — RAG Search, RAG Reindex
• Code — Run custom Python

Channels

A channel is where end-users interact with your deployed agents:

Deployments

A deployment publishes your agents or apps so they can serve live traffic.

• Preview — A safe sandbox for testing. Only your team can access it.
• Production — The live environment at yourapp.qraptor.app serving real users.

The deployment wizard lets you select agents, pin versions, configure schedules, and review credit costs before going live.

Tools & Integrations

Tools extend what your agents can do. qRaptor provides 40+ built-in tools (web search, database queries, document processing, vision, audio, and more) plus:

• MCP Servers — Connect to any Model Context Protocol server for additional tool access.
• Custom HTTP tools — Define your own API integrations per project.
• Agent-as-Tool — Call any Task Agent as a tool from within an AI Agent.

Remote Executors

Agents normally run entirely on qRaptor’s infrastructure. When a step needs to reach something that only exists inside your own network — a private database, an internal API, a GPU box — you can run a remote executor there instead.

It’s a lightweight agent you install on your own machine. It connects out to qRaptor and holds the connection open, so nothing on your side has to accept inbound traffic. You then mark individual code nodes, tools or MCP servers as Remote, and just those steps run on your infrastructure. See Remote Executors.

Versions

Every agent change creates a new version — an immutable snapshot of the graph, configuration, and tools. You choose which version to deploy and can roll back at any time.

How it all fits together

Workspace (billing + isolation)
└── Project (product boundary)
    ├── Agentic Workspace
    │   ├── AI Agents (conversational, autonomous)
    │   ├── Task Agents (workflow graphs)
    │   ├── MAS Systems (multi-agent coordination)
    │   └── Agent Personas & Tools
    ├── Apps (generated web/mobile/copilot apps)
    ├── Data Vault (tables, docs, content stores, memory)
    ├── Deployments (preview / production)
    ├── Configuration (LLM policies, settings)
    └── Integrations & Channels

Next steps

• Platform Overview — See the full platform at a glance.
• Create your account — Start building.