Skip to content
Mazen Chaban, home
Back to projectsCase study 03 / 16 · 2026

Personal tools · Artificial intelligence

MAZAGENT

Agents organised around the work.

An interface to assign a task, choose a role and find the agents’ work in one workshop.

Discuss a similar project
Sector
Personal tools · Artificial intelligence
My scope
UI adaptation · Instructions · Orchestration
Project status
Agent workshop · In use
Interface of the local MAZAGENT workshop, adapted from Rakazo.
Interface of the local MAZAGENT workshop, adapted from Rakazo.Enlarge the screenshot (new tab)
01 · Context

Organising the agents’ work.

I wanted to use my agents with precise roles: researching, writing, carrying out certain tasks and checking their results.

The project builds on Rakazo, an open-source base. My work covers adapting the interface, structuring the roles, the instructions and local routing.

02 · Design

Roles, tools, a check.

01

Give each agent a clear role

The interface shows what each role is for. A brief sets out the task and its limits before execution.

02

Separate production and verification

Distinct roles handle research, writing and checking. Local routing manages the models and their fallbacks.

How it works

Inside MAZAGENT.

  1. 01Brief
  2. 02Agent
  3. 03Execution
  4. 04Check
03 · Delivery

A local workshop, adapted from Rakazo.

The delivered scope

  • MAZAGENT interface: overview, search and work areas.
  • Guided brief and specialised roles for each task.
  • Shared instructions and separate working contexts.
  • Local router, execution checks and a dedicated virtual machine.

What it changes

Seventeen agents with defined roles: research, writing, execution, checking. Each one knows what it does, and what it does not.

Repetitive tasks run in the background. Actions with consequences wait for your approval.

What about your project?

Let’s talk about your project.

An application, AI agents, a business need. Tell me what you want to build.

WhatsApp (new tab) 06 25 66 66 75