Workflow automation
Turn repetitive intake, routing, and review into a clear, auditable flow.
Software engineering / MCP / AI integrations
I build tools for AI agents - and test what happens when they fail.
I'm Tomi Šeregi, a software engineer in Slovenia. My work covers MCP integrations, human-approval workflows, search infrastructure, and reproducible reliability tests.
Working code. Explicit permissions. Reproducible tests.
From agent request to inspectable boundary
01 — 03
Selected engineering work
Working systems, with evidence and limits.
Explore the implementation, what it demonstrates, and where its boundaries are.
DigestSEO MCP Suite
Four related search, data, and validation systems: mcp-geo, mcp-gsc, and mcp-web-validator have local and hosted/product surfaces; TrendPulse is community/self-hosted while its managed layer remains in development.
Nofax
Local human approvals, choices, refinements, and waits, paired with a deliberately narrower remote Worker for one-way notifications and safe request inspection.
Engineering noteEngineering note: recovering the newest pending request
A synthetic 21-record regression fixture shows why pending records must be ordered by durable creation time before a bounded result limit is applied.
What I build
Practical automation, shaped around the way your team already works.
The goal is not to add another tool. It is to remove a recurring bottleneck without giving away control.
AI-assisted operations
Use models where they help — drafting, classifying, and finding the next action — with people at the decision points.
Production systems
Ship the infrastructure, observability, and handoffs that make an AI workflow dependable.
Specialist services
Three areas I go deep on.
Each one is a standing offer with its own page, not a bullet on a capability deck.
01
AI enablement & compliance
AI literacy training and EU AI Act readiness for teams that have to answer for the tools they use.
Explore enablement →02AI visibility & content
Track how AI assistants cite you, and run the content pipeline that keeps you in the answer.
Explore visibility →03AI integration & reliability
MCP servers, evaluation harnesses, and the document and ERP plumbing underneath.
Explore integration →Products
Two focused products. One operating principle.
Useful automation should be easy to understand, easy to evaluate, and safe to own.
Engagements
Fixed fees, not hourly.
Send a short brief first. We can arrange a call if it helps clarify the scope. API and platform usage is billed at cost.
Automation teardown€500–900
AI Automation Auditfrom €2,500
Done-for-you Workflow Buildfrom €4,000
Care & Expansion Retainerfrom €800/month
How we work
A small first step, then a system you can own.
- 01
Map the bottleneck
Find the expensive, repeatable part of the work and define what good looks like.
- 02
Build the smallest useful flow
Prove the workflow with your rules, your data boundaries, and your team in the loop.
- 03
Operate and improve
Measure what happens in production, then expand only where the evidence says to.
Built for careful work
Trust is built into the workflow.
The right automation makes the next action clearer, while keeping responsibility visible.
Human control
Automation handles the first pass. Your team keeps the decisions that matter.
Clear boundaries
Data access, sending permissions, and escalation points are part of the design.
Measured in production
The system earns trust through observable, reviewable behavior — not a pitch deck.
Have a technical brief?
Send the project or repository, the result you need, and your deadline. A call is optional.
Send a brief