Agentic engineering for commerce

Ship commerce software faster, with AI agents inside your stack.

Senior engineers who put AI agents to work on Magento, Adobe Commerce and the systems around them — in your repositories, through your CI/CD, under your security rules.

Magendoo is a commerce engineering studio in Bucharest with 20+ years of production commerce systems behind it. Marketing ships without waiting on tickets. Engineering ships without fearing releases. Your environment, your code, your controls.

Magento / Adobe Commerce · Shopify · PrestaShop · Odoo · Go · Python · MCP · Reply within one business day
Proof

Built in public. Trusted in production.

What clients say about working with us — and the tools we use in delivery, published as open source by the same engineers who would work with you.

“Magendoo is really good — sharp and experienced, fluent in technical and business processes. They helped us move from a legacy system to a microservices architecture that handles meaningfully higher load at lower cost.”

CTO, Akoova · United Kingdom

“It was a pleasure to work with Magendoo. Their proactive approach and constant drive for the right solutions make them a valuable asset on any team. They delivered our payment integration ahead of schedule and under budget.”

Technical Lead, Buckaroo · Netherlands
Working parts we publish

MageMCP

Connects Claude or Codex to a Magento store through typed, permissioned tools. Reads are scoped, writes need confirmation, every action is logged. Your team asks for the report or the catalogue fix instead of waiting for a module.

Open source · 60 typed operations · 895 automated tests

Platform grounding

Agents work from the platform's real documentation and your engineering standards, not from guesses. We built this layer for Magento, Adobe Commerce, PrestaShop and Shopify for our own delivery work, then published it.

Magento 2 Extension Atlas

A searchable catalogue of public Magento 2 modules, themes and projects with activity in the last two years, refreshed daily. See what already exists — and what is still maintained — before you build or upgrade.

AI-assisted test generation

An implementation note from an anonymised Magento 2 retail project: where AI-written tests fit next to human review, and where to be careful.

The bottlenecks

Where commerce engineering stalls — and what we change.

Four patterns slow down mid-market commerce teams. None of them is a staffing problem. They are workflow problems, and workflow is what we change.

Marketing and reporting wait on developers.

The banner swap, the campaign page, the “can you pull the numbers” request: each one becomes a ticket behind the feature work.

What we change Typed, permissioned tools on your store. Your team asks for the report or stages the change; reads are scoped, writes need confirmation, every action is logged.

Releases are events, not routine.

Maintenance windows, peak-season code freezes and manual QA, while the backlog keeps growing.

What we change Agents write tests and open pull requests. Your CI, your static analysis and your reviewers decide what merges.

Integrations fail quietly.

Store, ERP and fulfilment drift apart, and the first alert is a customer asking where their order is.

What we change Integration workers in Go or Python that log, retry and reconcile, so a failure shows up in your monitoring, not in a customer's email.

Security blocks AI adoption.

Developers want Claude Code or Codex. Security will not send source code or customer data to someone else's sandbox.

What we change The workflow runs inside your perimeter: your repositories, your approved model provider, your data boundaries.

Engagements

Two ways to work with us.

Agentic Engineering Assessment

A code-level look at how you build software, and a plan for agents inside it.

Fixed fee, agreed before we start · typically 1–2 weeks
You receive
  • Agentic Engineering Scorecard
  • Workflow and architecture proposal
  • Prioritized backlog
  • Security and governance model
  • 90-day implementation plan
  • Expected measurable outcomes

Embedded Agentic Engineer

A senior commerce engineer inside your team, using agents to ship.

2 days/week · 3 days/week · dedicated
What it covers
  • Implementing the assessment roadmap
  • Agentic workflows your team keeps
  • Production software, shipped along the way
  • Monthly objectives and measurement
  • Where you have no AI lead: architecture, governance, vendor selection

Not sure which fits? Describe the problem. We will tell you which one fits — or that neither does.

How we work

Agents come to your environment. Not the other way around.

We don't ask you to hand your codebase, credentials or customer context to someone else's sandbox. We establish the agentic workflow inside your perimeter:

Your repositories
GitHub, GitLab, self-hosted
Your CI/CD and deployment process
agents open PRs; your pipeline decides
Your model provider
Anthropic, OpenAI, AWS Bedrock, Google Vertex, or your gateway
Your standards
PHPStan levels, coding standards, SonarQube, Snyk, Semgrep, coverage thresholds
Your approval rules
PR-only, required reviewers, protected branches
Your data boundaries
no source or customer data leaves infrastructure you control

If your rules say two approvals and no direct commits, those are the agent's rules too.

Field notes.

→ All insights
Questions

Before you write to us.

Agentic engineering is the discipline of putting AI agents to work on real software — planning, writing, testing and reviewing code under the same controls human engineers follow. Applied to commerce, it covers two things: using agents to build and maintain your store's software, and building agents that run parts of the operation itself.

Magento and Adobe Commerce are our home ground. We also work on Shopify, PrestaShop and Odoo, and build the services around them in Go and Python: integrations, background workers and MCP servers.

You send a few lines about what you want to fix or ship. An engineer replies within one business day, from [email protected]. If there is a fit, a short scoping call follows. No sales team, no email sequences.

Every engagement runs on a fixed fee, agreed before we start — never hourly, never open-ended. We quote it after the scoping call, once we know the platform, the repositories and the goal.

No. Agents open pull requests into your process, and your developers or your agency review and merge them. The embedded engineer works inside your team, not around it, and the workflow stays with your team.
Start a conversation

Tell us what you want to fix or ship.

An engineer reads every message and replies within one business day, from [email protected]. No newsletters. No sequences.

Based in
Bucharest, Romania · working across EU time zones
Paperwork
NDAs and security questionnaires — send them along.
Prefer a structured request? Request an assessment →