Skip to content

Martin Gratzer · Staff+ Engineer · Remote, worldwide

Software that holds up in production.

I've spent over 20 years building software and shipping it with teams. Agents are now a regular part of that work. I usually come in when things get tricky: for a project, or as a fractional Staff+ engineer.

You work directly with me.
Martin Gratzer, founder of topmind GmbH

Where I usually help
  1. An important project needs to shipExtra experience on the team gets it there faster, with less risk.
  2. A decision with real consequencesArchitecture, build vs. buy, migration: worth getting right the first time.
  3. A system that grew over yearsIt has to keep running while it changes.
  4. Agents that are more than an experimentIn the product or the development process, with a clear benefit.

Services

Two ways to work together.

Project

Software engineering

I build systems end to end and take them to production.

Product features
With an eye on product and users, from idea to release, right inside your existing product.
Modernizing existing systems
Rebuilt step by step while everything keeps running.
Agentic systems
Agent features inside your product, and MCP servers that let agents use it. Where it actually makes sense.

A good fit if you have a concrete project and need someone to build it.

Format: a project with a clear goal

Discuss a project

Ongoing

Fractional Staff+

Staff+ experience when you need it, without a full-time hire.

Architecture decisions
Build vs. buy, migrations, tech debt. Decisions that don't haunt you later.
AI strategy
Working with leadership on where AI actually pays off, then guiding the way there: in the product, the team and the interfaces.
Team enablement
Supporting your teams day to day, including how they work with agents: workflows, workshops, pairing.

A good fit if your team needs Staff+ experience but has no full-time role for it.

Format: ongoing, a few days a month

Discuss a fractional role

AI Enablement

Agents in three places.

I work with agents every day, on real projects. Together we figure out where they help your team, and where they don't.

Agents that work reliably inside a product need more than a good model.

  1. 01

    Clear value

    Agent features for concrete tasks your users have, not a chat window for everything.

  2. 02

    Proper permissions

    Agents act on behalf of your users, with exactly the rights they need for the job.

  3. 03

    Evals and traceability

    Quality gets measured, not guessed, and every step an agent takes can be traced.

  4. 04

    Swappable models

    An architecture that doesn't lock you into one provider.

Not every feature needs an agent.

Often the simpler solution is the better one. When that's the case, I'll tell you.

From issue to merge: who does the work at each step.

  1. 01

    Issue

    Goal, context and constraints

    Human

  2. 02

    Plan

    Agent proposes, team approves

    Agent + human

  3. 03

    Code

    Implementation in your codebase

    Agent

  4. 04

    Test

    Verifies itself against your checks

    Agent

  5. 05

    Review

    Agent pre-reviews, team decides

    Agent + human

  6. 06

    Merge

    By the rules you set

    Your call

Together we decide where humans stay in the loop, based on your team and your risk.

  • Workflow design

    Skills, conventions and guardrails for your codebase and your issue tracking.

  • Workshops & pairing

    Your team works on its own code and real tickets, not on demos.

  • Workflows that get better

    What matters isn't how often agents get used, but whether each run makes the next one easier.

ForgeOpen source

The workflow I build with every day: agent skills that carry a task from issue to merge.

Many people now do their work inside an AI assistant. This is how your product becomes usable there.

  1. 01

    MCP servers

    Your product as a tool that agents can call directly.

  2. 02

    Secure sign-in

    OAuth instead of shared keys, with clear rights for every access.

  3. 03

    Interfaces for agents

    APIs and command-line tools that agents understand and use reliably.

  4. 04

    WebMCP

    Open your web app to agents right in the browser, without a server of its own.

Your Website Is the MCP Server NowArticle

How I gave a web app an interface for agents over a weekend, with no server and no auth to build.

Read the article

Process

How we start.

  1. 01 · Conversation

    30 minutes, no strings attached.

    You tell me where you stand. I'll tell you honestly whether and how I can help.

  2. 02 · Assessment

    Clarity before code.

    Where it actually hurts and what matters first. With or without agents.

  3. 03 · Delivery

    Build, support, hand over.

    As a project or as fractional Staff+, so your team can keep going on its own afterwards.

Portrait of Martin Gratzer, founder of topmind GmbH

About

Martin Gratzer

Staff+ Engineer, builder and consultant · Founder of topmind GmbH

For a long time I mostly built native iOS and Android apps, later full-stack systems and architecture across teams. I've always worked closely with product people and helped shape features, not just build them. I've taken products from the first line of code to millions of users, and seen pretty much every mistake you can make in production. I made some of them myself.

These days I'm focused on how humans and agents build software together. That's new for all of us. What stays are the fundamentals: clear decisions, simple systems and teams who understand their work.

Let's talk.

Want to build with agents, or need a second opinion on where AI fits in your team? I'd be happy to talk.

Book a time

Email

martin@topmind.eu

Phone

+43 680 233 9000

Mon–Fri, 9am–5pm CET