Skip to main content
SCENARIOS

What agent teams do in real work

Six ways to run multi-model (Claude · Codex · Grok) agent teams in a Marblo workspace.

Scenarios to validate in a four-week pilot. If one fits your workflow, we can review it in the free first assessment.

How it's used

Three ways to use Marblo

Build a team by role, connect workflow steps, and check results on the board. Evaluate adoption through assessment, training, and a pilot.

Role-based agent teams

A pattern for sharing work among Claude, Codex, and Grok agents assigned by role.

  • Role-based independent agents
  • Optimal model per task
  • Manage subscriptions and usage limits per account
Agent flows

Automate complex workflows with node-based pipelines (LLM→Agent→Human→Output).

  • Node-based flow builder
  • Human review along the way
  • Repeatable execution & flow saving
Workspace task management

Follow agent tasks and provide feedback through the Marblo kanban board and terminals.

  • Kanban board + real-time management
  • Code viewer + multi-terminal
  • PM feedback sent to agents in real time
SCENARIOS

Scenarios by Function

Scenarios for using Marblo agent teams across workflows. Confirm the scope that fits your team through an assessment and a pilot.

Task Automation Agent

Handles repetitive tasks automatically as part of the agent team.

  • Document creation and data entry automation
  • Schedule management and reminder sending
  • Report generation and analysis
Marketing Automation Agent

As part of the agent team, automatically plans and executes personalized campaigns.

  • Customized content creation per customer segment
  • Email/SMS campaign automation
  • Performance analysis and optimization suggestions
Sales Support Agent

As part of the agent team, supports the entire sales process from leads to deal closure.

  • Lead scoring and prioritization
  • Automatic follow-up and consultation booking
  • Proposal and quote generation
Customer Service Chatbot

As part of the agent team, responds instantly to customer inquiries 24/7.

  • Natural language understanding and context awareness
  • Multi-language support (KR/EN/JP)
  • Complex inquiries transferred to agents
Internal Document RAG System

As part of the agent team, searches internal guides to answer questions that need specialist knowledge.

  • Vectorization and search of vast internal documents
  • Accurate source citation and references
  • New documents added and kept up to date
Data Analysis Agent

As part of the agent team, analyzes data and automatically derives insights.

  • GA4, BigQuery data analysis
  • Trend and anomaly detection
  • Automatic report and dashboard generation
Workflow

How Work Flows in Marblo

Split work, assign agents, and have a person review the results. See the enterprise adoption ladder for the adoption process.

  1. 01

    Break work into tasks

    Break the goal into tasks and map how they depend on each other (a DAG)

  2. 02

    Agent Team Design

    Assign optimal models (Claude, Codex, Grok) by role

  3. 03

    Flow Setup & Testing

    Connect the work steps in the flow editor and test with real scenarios

  4. 04

    Review and approve results

    Your team reviews the results in the workspace and approves them

How work is assigned and reviewed

Input → Orchestrator → Agent Team → Human review → Output

Stack

Tech Stack

What you use to run agent teams in the Marblo workspace. Build technologies for each scenario are listed on that scenario page.

Agent
Claude Code (Anthropic)
Agent
Codex (OpenAI)
Agent
Grok (xAI)
Protocol
MCP (Model Context Protocol)
Platform
Marblo
Automation
Flow Editor

Core Features

  • Each agent has its own account and usage limit
  • Each task gets the model that fits it
  • A flow editor that connects work steps as blocks
  • Real-time kanban board
  • Review comments reach the working agent right away
  • Published transfer scope · local execution
FAQ

Frequently Asked Questions

How does adoption work?
Free first assessment (1–2 hours, one-page summary) → AX assessment (half day + 1-week report) → team training (2 hours) → 4-week pilot. One team runs Marblo on real tickets, and the pilot ends with before/after numbers. The results inform the Marblo license decision.
What tech stack do you use?
Agents such as Claude, Codex, Grok run in a Marblo workspace, with MCP connections to filesystems, databases, APIs, and Git. The AX assessment checks the required connections and the pilot verifies them.
Can it integrate with our ERP, CRM?
Yes. MCP (Model Context Protocol) connects it to filesystems, databases, APIs, Git, and most internal systems.
How much does it cost?
The free first assessment takes 1–2 hours and includes a one-page summary. AX assessment, team training, and pilots are quoted after consultation; credit terms are explained with the quote. Marblo licenses follow public pricing on marblo.app.
Who runs it after the pilot?
Operations and company-wide rollout are yours. The pilot leaves your team the verified operating rules and the training to run them; we supply the product as Marblo licenses. We do not sell managed operations.
Next

See a pattern that fits? Verify it in a pilot

The AX assessment sets the scope and metrics; the four-week pilot ends with before/after numbers. AX assessment, team training, and pilots are quoted after consultation, with credit terms explained in the quote.

Marblo Scenarios | Six AI Agent Team Patterns | Hypemarc