GitHub
AI-Native Business Operating System

Brezel.io

The ERP that builds itself and employs AI like people.

ERP coreAI agentsShared rightsHuman oversight

Describe the outcome. Brezel handles execution across your entire business operations and enterprise workflows— from customer support to procurement to compliance Brezel doesn’t just automate work. It learns how your company works

Be the first to try → View on GitHub
01Shared rights system
02People and agents
03ERP + orchestration

Where Automation
Becomes Infrastructure.

Design workflows visually. Brezel runs them with modular AI agents. Scale across your entire business operations without writing a single line of code.

How it works

From Signal to Learning.

Step 01
Signal
Anything can be a signal: an e-mail, a webhook, a document or a request
Step 02
Case
The signal is classified, checked for relevance and turned into a case
Step 03
Decision
The right person approves, edits or rejects the proposed next step
Step 04
Action
An agent or employee replies, records the outcome or builds what is needed
Step 05
Learning
The outcome feeds role memory and playbooks as the learning loop develops
In practice

A wedding catering inquiry arrives by e-mail. Brezel opens a case, sees that date, guest count and budget are missing, and drafts the follow-up. The owner approves with one click

Governance & security

Governance built in, not bolted on

Designed for the EU AI Act era: humans stay accountable for every decision that matters

01

Roles & permissions

One rights system for employees and agents. Agents act only within their department

02

Secrets stay secret

An integrated password manager uses credentials at runtime; they never enter the LLM context

03

Sandboxed work

Agents use isolated execution environments for code, shell and tools, away from production systems

04

Fully traceable

The signal pipeline records what came in, what was decided and by whom

05

Multi-tenant

Several companies on one installation, with separate data, users and agents

06

Cloud or on-prem

Run on-premise with local LLMs or use Brezel Cloud when available. The model is chosen per agent

Self-building ERP

No agency project. Brezel builds the system from your goals and your inbox

Start with the company you have. Extend the ERP as real work reveals what is missing

01 / SETUP

Describe your company. Get your system

Brezel proposes the modules a business like yours needs and builds them once you confirm. Module suggestions in the setup wizard are in progress

Example · a bakery
CustomersInquiriesOrdersInvoicesSuppliersWeekly Planning
02 / EVERYDAY WORK

Real signals reveal what is missing

In daily work, an incoming document can reveal the need for a new module

  1. 01An invoice PDF arrives as a signal
  2. 02Brezel sees there is no invoice module
  3. 03It drafts fields, layout and workflow
  4. 04You review the preview and approve
  5. 05The module goes live and the invoice is filed

The system grows with the company instead of being configured once

Human in the loop · planned dashboard

People only see what needs them

Brezel works in the background. The planned decision dashboard will surface only the decisions that need a human

Approve · Edit · Reject. Each choice teaches Brezel.

The full signal pipeline stays available for admins and audits

Powerful Features

Everything you need to build, automate, and scale workflows with Brezel.
Permissions
Workflows
Notification Center
The Many Uses of AI
Modular AI Agents, Not Monolithic Code
Revision History & Comments
Smart Document Management
Centralized Email System
Flexible Data Output
Feature

Permissions

Define precise access control across modules with full read, write, update, and delete permissions.

Ensure sensitive data stays protected while giving each user exactly the access they need.

Permissions
Core Concept

Company OS
Agentic Policy

Defines how employees, AI agents, and management interact — ensuring decisions, communication, and collaboration are efficient, transparent, and ethical.

01 AI Mirror Agents

AI Mirror Agents

Each employee can have an AI agent that acts as their mirror — carrying the same permissions and working as their intelligent proxy. For privacy, employees communicate with other employees only, never with their colleagues' mirror agents.

  • Chat with your mirror agent
  • Assign it tasks directly
  • Let it make decisions on your behalf (with approval)
02 Transparency

Transparency

All interactions with independent agents are fully traceable. Management can review every action at any time — nothing operates in the dark.

03 Communication Rules

Communication Rules

Hierarchy-aware boundaries built on trust and responsibility — not rigid technical locks. Employees interact freely with agents below them; reaching across or upward requires approval from a higher-level human, scoped to that specific task or case.

04 Independent Agents

Independent Agents

Agents without a human mirror. They handle general system tasks — monitoring, content creation, workflow automation — acting only on behalf of management or the employees who activate them.

05 Custom Policies

Custom Policies

Beyond the defaults, every company can define its own access rules — granting full, limited, or no access to specific agents per employee.

Full accessLimited accessNo access
Get Started

See Brezel in Action

Experience how Brezel transforms workflows into structured, autonomous systems. No guesswork. No manual coordination.

Be the first to try → View on GitHub