Skip to content
VZU
VZU OS · The 12 agents

Twelve specialized agents on a single runtime.

Atlas, Vega, Kai, Nova, Pixel, Echo, Sketch, Launch, Sentinel, Oracle, Quill, Hunter. MCP-native. RBAC at the agent level. Audit trail from day one. The VZU Agentic OS is the engine that ships the work.

Open the runtime Browse the marketplace

✓ Reply within 1 business day · No credit card · No demo gate

Scroll

3 agents

Engineering

Build · Data · Security · Front-end · Back-end · Deploy

5 agents

Design

Visual contrast · Layout · Motion · Token systems

2 agents

Review

Security review · QA · Observability

1 agents

Data

Long-form writing · Document automation

1 agents

Automation

Lead routing · Account research · Outreach

All 12 agents

Every agent is a real, working component of the runtime.

Agent · 01

ATLAS

Full-stack app builder.

engineering · VZU Build

Toolset

  • → mcp.file.read
  • → mcp.data.schema
  • → mcp.code.generate
View agent trace

Agent · 02

VEGA

React/Next.js engineer.

engineering · VZU Web

Toolset

  • → mcp.code.read
  • → mcp.code.write
  • → mcp.test.unit
View agent trace

Agent · 03

KAI

Backend + infra engineer.

engineering · VZU Stack

Toolset

  • → mcp.db.migrate
  • → mcp.api.deploy
  • → mcp.queue.worker
View agent trace

Agent · 04

NOVA

Editorial UI designer.

design · VZU Surface

Toolset

  • → mcp.figma.write
  • → mcp.code.write
  • → mcp.a11y.check
View agent trace

Agent · 05

PIXEL

Motion + WebGL designer.

design · VZU Canvas

Toolset

  • → mcp.three.scene
  • → mcp.shader.compile
  • → mcp.motion.choreograph
View agent trace

Agent · 06

ECHO

WCAG accessibility auditor.

design · VZU Design

Toolset

  • → mcp.a11y.scan
  • → mcp.screenshot.compare
  • → mcp.report.write
View agent trace

Agent · 07

SKETCH

Wireframe + prototype designer.

design · VZU Sketch

Toolset

  • → mcp.figma.wireframe
  • → mcp.figma.prototype
  • → mcp.figma.handoff
View agent trace

Agent · 08

LAUNCH

Landing page composer.

design · VZU Launch

Toolset

  • → mcp.site.build
  • → mcp.motion.choreograph
  • → mcp.aeo.schema
View agent trace

Agent · 09

SENTINEL

Code reviewer + security auditor.

review · VZU Audit

Toolset

  • → mcp.code.read
  • → mcp.semgrep.run
  • → mcp.owasp.scan
View agent trace

Agent · 10

ORACLE

Production prompt engineer.

review · VZU Prompts

Toolset

  • → mcp.prompt.write
  • → mcp.eval.run
  • → mcp.prompt.iterate
View agent trace

Agent · 11

QUILL

Spreadsheet engineer.

data · VZU Sheets

Toolset

  • → mcp.xlsx.read
  • → mcp.xlsx.write
  • → mcp.xlsx.reconcile
View agent trace

Agent · 12

HUNTER

Resilient web scraper.

automation · VZU Scrape

Toolset

  • → mcp.scrape.run
  • → mcp.proxy.rotate
  • → mcp.robots.check
View agent trace

Orchestration

Twelve agents, one orchestrator, one audit trail.

The NetWit runtime orchestrates the 12 agents. The orchestrator is the supervisor; the agents are the workers. Every message is logged. Every state is recoverable. Every action is auditable.

  • A → B → C
    01

    Relay

    active

    Sequential hand-off. Each step waits for the prior step to complete. The simplest pattern. The most auditable.

    1. 1. Brief → A
    2. 2. A → result
    3. 3. result → B
    4. 4. B → result
    5. 5. result → C
    6. 6. C → final

    Use case

    Brief: build the schema, then build the API, then build the UI. Atlas (schema) → Kai (API) → Nova (UI).

  • A + B + C
    02

    Parallel

    active

    Independent subtasks run in parallel. Results combined by the orchestrator. Used when the brief splits cleanly.

    1. 1. Brief → orchestrator
    2. 2. orchestrator → [A, B, C]
    3. 3. A → result
    4. 4. B → result
    5. 5. C → result
    6. 6. orchestrator → combine

    Use case

    Brief: refactor backend, frontend, and audit at once. Kai + Nova + Sentinel in parallel. Combined by the orchestrator.

  • A → {B, C, D} → A
    03

    Supervisor

    active

    A delegates, observes, re-delegates. The most powerful pattern. The most audited.

    1. 1. Brief → A (supervisor)
    2. 2. A → B, A → C, A → D
    3. 3. B → result
    4. 4. A reviews B
    5. 5. C → result
    6. 6. A reviews C
    7. 7. A → re-plan if needed
    8. 8. A → final

    Use case

    Brief: build a multi-tenant system with auth, billing, and observability. Atlas (supervisor) → {Kai, Vega, Sentinel}. Atlas re-plans until the brief is met.

Run the runtime

See all 12 agents in motion.

The runtime demo shows the orchestrator, the agent registry, and the live execution trace. Every agent has a live playground at /vzu-os/agents/<slug>/playground/.

Open the runtime Browse the marketplace