A frontend for humans. An interface for agents.
We build the machine-readable layer AI agents need to understand, trust and act on your offering.
A new class of visitor
Alongside humans, AI agents increasingly arrive on your site on behalf of real users. They don't browse like people: no images, no hero section. They parse structure, meaning and available actions, and decide in seconds whether your site is a reliable source.
Built for humans
- Visual layout, images, scroll
- Content only after a click or interaction
- Logic implied in the design
- Actions through UI flows
What agents need
- Clear semantics and structure
- Content in the DOM, machine-readable
- Verifiable, canonical facts
- Permitted actions, cleanly defined
From user to agent to your offering
- 01
Someone asks an agent
A user has ChatGPT, Perplexity or Claude research, compare or handle something you offer.
- 02
The agent reaches you
Instead of clicking, the agent parses your site. It needs structured data and clear actions rather than a visual layout, and gives up fast when it finds neither.
- 03
The layer answers
The Agentic Layer hands it readable facts, trust signals and prepared actions, from an inquiry to a booking or a purchase.
A layer alongside your existing website
The website stays for humans. The Agentic Layer adds an agent-facing layer that follows from clean work and improves classic SEO and accessibility at the same time.
Semantic Map
Who you are, what you offer and for whom, mapped machine-readable.
Structured Content
Products, services, prices and terms as structured data.
Action Model
What an agent may prepare: an inquiry, a comparison, a cart, an appointment.
Trust Model
Canonical facts, policies, proof and freshness an agent can rely on.
API / MCP Endpoint
Structured endpoints instead of scraping, so agents query verifiably and current.
Agent Instructions
Rules, not just data: what is reliable, what needs approval, when to defer to a human.
How an agent queries your layer
An agent asks in natural language, queries our own layer in a structured way, and gets machine-readable answers plus the next actions. Real data from /api/agent/workshops.
User to their agent
Which workshops are bookable, and what do they cost?
GET /api/agent/workshopsAnswer, structured and machine-readable
The Concept
- price
- 199 EUR
- format
- online
- tier
- 1
Hands-on
- price
- 599 EUR
- format
- online
- tier
- 2
Team Enablement
- price
- 4900 EUR
- format
- hybrid
- tier
- 3
Actions the agent can prepare
Live from /api/agent/workshops, the REST mirror of our MCP endpoint.
For websites
- Understand the service
- Find the right page
- Prepare an inquiry
- Generate a briefing
- Answer FAQs
- Read trust signals
For eCommerce
- Find products
- Compare variants
- Check availability
- Prepare a purchase decision
- Prepare a cart
- Hand checkout to a human
From audit to a layer that runs
Clearly scoped steps, senior-led, integrated into your stack.
Agent Readiness Audit
A score across six dimensions, quick wins, an honest read.
Agent-facing content model
Data model, trust model and action map for your product.
llms.txt & agent.json
Manifest and instructions that give agents an entry point.
Structured data
Schema cleanup and machine-readable content in the DOM.
API / MCP layer
Structured, verifiable endpoints, integrated into CMS, shop, PIM or ERP.
Monitoring & iteration
Re-scans, agent simulations and freshness as an ongoing operation.
We run the layer on ourselves
This site has a public agent endpoint: agents can query our services and workshops in a structured way and prepare a booking, while a human closes it. Internally we deliver AI-native product development, design systems and a workflow with many AI agents, with reference contexts from eCommerce and enterprise.
Externally confirmed
Google now documents it too: AI agents read websites via the DOM, accessibility tree and screenshots, and need agent-friendly structures. That is exactly what we measure and build.
Google's AI optimization guide