Blog · Published · Updated
An agent-ready company website: what we built on syntalith.ai
Short answer
In October 2026 we released seven parts on syntalith.ai for AI assistants and agents: from robots.txt and an API catalogue to an MCP server with our services and available times, and WebMCP on the booking page. An agent can read our services with prices and prepare the booking form, and a person always confirms the meeting.
How we workDetails ↓
- 01Call30 minutes with an engineer, free
- 02PrototypeA working system on your process and data, free
- 03ImplementationTarget and fixed price in the contract; 50% at signing
- 04Acceptance50% once the agreed target is met
- net per application, fixed price before contract
- from €3,000
- from signing the contract to acceptance
- 4–10 weeks
- free prototype on your documents
- €0
We belong to the OpenAI and Anthropic partner programmes and learn through their official courses. What this means ↓
Author: Syntalith · 3 min read
Statistics
Websites and AI agents: statistics 2026
7
parts for AI agents released on syntalith.ai in October 2026: 6 read-only and 1 with a person's confirmation at booking.
8
According to the A2A Protocol project (2026), version 1.0 of A2A was released on 12 March 2026, and the standard is led by a technical committee with representatives of 8 technology companies.
24 March 2026
According to OpenAI (2026), from 24 March 2026 the ACP protocol supports product discovery in ChatGPT, and merchants send product feeds and promotions through it.
Source: OpenAI, Powering Product Discovery in ChatGPT (2026) ↗
34.7%
According to Eurostat (2025), 34.7% of EU enterprises that used AI used it for marketing or sales.
Figures checked: 22 October 2026
What we released
| Part | Address | What it gives an agent |
|---|---|---|
| Signals for bots | /robots.txt | permission to search and to use the content in AI answers |
| API catalogue | /.well-known/api-catalog | a list of public interfaces in the RFC 9727 format |
| Service catalogue | /api/agent/catalog and /openapi.json | services, starting prices and terms in JSON with an OpenAPI description |
| Guide for agents | /llms.txt | a table of contents and the access documentation |
| Markdown versions | for example /llms/en/pricing.md | the page content without navigation or scripts |
| MCP server | meet.syntalith.ai/mcp | reading services, evidence and available times |
| WebMCP at booking | the booking page | preparing a visible form in the browser |
Pages that models can read
Every important page has a Markdown version. When an agent sends a request with the Accept: text/markdown header, the server sends it to that version, while an ordinary browser gets HTML with the Markdown address in the Link header. Prices are in the text and in the structured data, with a starting price, a fixed price and a quote told apart. An agent comparing suppliers gets the same amounts as a person: automations, applications and agents from €3,000 net, training from €1,150 net a day.
- 01Readingservices and prices in Markdown or JSON
- 02Evidencematerial that shows our experience, through the MCP server
- 03Timesavailable call slots, read only
- 04FormWebMCP fills in the visible booking form
- 05Confirmationthe user sends the form and confirms the meeting by email
What we left out on purpose
- No A2A agent: at this stage a service company does not need an agent that negotiates with other agents
- No separate section for agents: people and agents read the same content
- No booking without confirmation: a person approves the slot in the calendar
Booking runs through a separate interface with approved access. Our rule is that an agent may prepare everything, and a person approves the commitment. Shopify uses a similar model in checkout, as described in the article on Shopify WebMCP. We do not report sales results yet, because the changes have been live since October 2026.
An MCP server with our services and available times
The public MCP server lets an assistant read the list of services, material that shows our experience and the available call slots. It is read only, so an agent cannot change anything in our systems. Booking is a separate interface with approved access, and the participant confirms the meeting by email. We recommend the same split to clients: reading for any agent, and action after a person's consent. An agent comparing suppliers for a buyer gets exact prices and evidence without going through menus, and the buyer sees in the form exactly what the agent prepared.
Where to start in your own company
Prices and services only in a PDF or an image
move them into the page text and structured data
audit from €600
AI search bots blocked in robots.txt
permission for bots and measurement of visits from assistants
audit from €600
Content and measurement work
an MCP server with your services and available times
from €3,000
Check in robots.txt whether AI search bots may read the site; OpenAI's bot documentation lists their names. You can measure visits from assistants with a parameter, for example utm_source=chatgpt.com. We build interfaces for agents as part of AI agents, and we check the website in an audit. For the free analysis, the website address and a description of the services an agent should find are enough.
Worked example · Scenario
How we count the payback: a service company
We run the same count for your company in the free analysis.
€4,363
a month for handling 1,500 job emails
Payback: 0.8 mo.
5 minutes to read an email and reply and checking in 15% of the old time are our assumptions. €34.90 an hour is the EU average hourly labour cost for 2025 according to Eurostat, for companies with 10 or more employees. Model cost: Claude Sonnet 5.5, about 3,000 + 600 tokens per message with a draft reply, at the ECB rate of 1.1205 USD per euro on 22 October 2026.
See the full calculation →
