
Integrating Exact Online With Your Own Software: When, How and What It Costs
A practical guide to an Exact Online integration: when it pays off, what is technically possible, the three hardest pitfalls and what an integration costs.
Insights, updates, and practical thinking on building modern workflows, automating execution, and using AI to reduce operational friction.

A practical guide to an Exact Online integration: when it pays off, what is technically possible, the three hardest pitfalls and what an integration costs.

An honest view of custom software maintenance costs: the 15 to 20 percent rule of thumb, what maintenance covers, and what happens if you skip it.

Mollie or Stripe for your webshop, SaaS or platform? A practical comparison on iDEAL, subscriptions, international growth and marketplaces, from our build practice.

In 2026 AI agents go beyond a chatbot: they perform tasks inside your systems. What makes an agent different from a chatbot, why MCP and context engineering are the turning point, and how to have a reliable AI agent built for your business processes.

In 2026 several Dutch subsidies partly cover custom software and digitalisation, from the Sprint subsidy to the JTF scheme and the WBSO tax incentive. Which schemes cover software, how a digitalisation advisory works, and how to set up a subsidy-eligible project.

The Dutch Cybersecurity Act (the national NIS2 implementation) is expected in Q2 2026. What it means for custom software and suppliers: supply-chain responsibility, duty of care, the 24-hour reporting obligation, and how to build NIS2-ready software now.

On 2 August 2026, the heaviest part of the EU AI Act becomes enforceable. What it concretely means for Dutch and European SMEs using AI in software: the seven documentation requirements, the provider vs. deployer split, the fines, and a practical checklist.

Headless AI shifts software from screens to actions. Learn how companies in 2026 build agent-ready APIs, MCP servers, audit trails and human-in-the-loop workflows.

A practical case study on Google’s 2026 spam update, pSEO risk, boilerplate content, structured data and the changes we made to make scalable content more useful.

Old systems do not break loudly. They bleed slowly through outages, manual workarounds, and lost deals. Here are seven signals that your legacy software is costing more than replacement, and what to do about each one.

Most agencies look identical in the sales pitch. These 12 questions separate the agencies that ship working software from the ones that burn your budget. Use this checklist before signing any contract.

Off-the-shelf SaaS gets you running in days but limits your competitive edge. Custom software fits perfectly but takes time and budget. Here is the decision framework we use with every client.
12 of 70 articles
This page lists 70 articles on custom software, AI tooling and development workflows. They are written from inside live projects at MG Software in Haarlem, not from research papers or hyped twitter threads. Each piece starts with a concrete question we recently hit in a client call or pull request, and ends with a decision we actually shipped (including trade-offs that turned out worse than expected).
Current breakdown: 18pieces on AI & automation (Cursor, Claude Code, MCP, agentic coding), 25 engineering deep-dives (Next.js 16, Postgres, Supabase, Three.js patterns), 11 workflow guides for founders and PMs, and 16 updates on Refront and client work. Filter via the tabs above, or subscribe to the RSS feed to get each new post automatically. No newsletter popups, no tracking pixels, no affiliate links.
Every post is written by Jordan Munk or Sidney de Geus, both co-founders of MG Software. We publish around two pieces a week and stick to three rules: the topic comes from a real project (not from an AI-generated topic list), numbers and screenshots are our own material or linked to a primary source, and when we are uncertain about something we write that down instead of implying a stronger claim than we can defend.
The content tracks what we build day to day. In Q4 2025 our stack moved from the Vercel AI SDK to Anthropic Claude Code for server-side agents, and that transition is documented in a series on agentic patterns, MCP tooling and code review. For productive refactors we have used Cursor 2.0 with composer-2-fast as the default model since early 2026. What we learn in those migrations lands here directly, so you can read along before you need to make the same call.
Missing a post on a specific tool, framework or architectural decision? Email info@mgsoftware.nl with the question or frustration, and if it fits our experience we will add it to the queue. Posts are never bought or sponsored; we can only write about what we have built ourselves or evaluated thoroughly.


















Describe your idea and receive a free analysis from our developers within 24 hours.
Get in touch and discover how MG Software can transform your ideas into working software.