yes, this page is for the robots
Built for agents.
Most websites make an AI agent guess: read the pixels, fake the clicks, hope the form submits. This one doesn't. Aahfil publishes a typed tool surface, clean markdown of every page, and an open door to every major AI crawler.
https://aahfil.com/api/mcpThree ways in
MCP server
A Model Context Protocol server over Streamable HTTP. Ask what we do, check when we're free, score a business's marketing maturity — as typed tool calls, not scraped HTML.
Markdown mirrors
Append .md to any page, or send Accept: text/markdown. You get the content without the layout, at roughly a tenth of the tokens.
Open crawling
Every major AI crawler is explicitly allowed in robots.txt — GPTBot, ClaudeBot, PerplexityBot, Google-Extended and the rest. Read us, cite us, quote us.
Connect
Any MCP client that speaks Streamable HTTP can connect directly. No API key, no signup.
In Claude, add it as a custom connector with that URL.
To poke at it from a terminal:
npx @modelcontextprotocol/inspector https://aahfil.com/api/mcpWhat the server can do
| Tool | What it does | Writes? |
|---|---|---|
| list_services | Every service we offer, with a one-line summary and a link. | No |
| get_service | Full detail on one service: scope, deliverables, approach, FAQs. | No |
| get_agency_info | Who we are, where we are, how to reach us, how we run projects. | No |
| check_availability | Open intro-call slots for a given date, Kuwait time. | No |
| run_growth_scorecard | Scores a business's marketing maturity 0–100 and returns prioritised fixes. | No |
Booking through an agent is currently switched off — send people to the contact page instead. It gets enabled once we're happy with what the logs look like.
Markdown, two ways
Both return the same content. The canonical URL is always the HTML page — the markdown is a representation of it, and says so in a Link header.
curl https://aahfil.com/en/services/seo.mdcurl -H "Accept: text/markdown" \
https://aahfil.com/en/services/seoMachine-readable files
- /llms.txtA short index of the site, written for language models.
- /llms-full.txtEvery URL we publish, in one file.
- /.well-known/mcpServer discovery manifest — capabilities before connecting.
- /.well-known/mcp/server-card.jsonServer card: what this server is and what it exposes.
- /sitemap.xmlEvery page, both languages, with hreflang.
- /robots.txtCrawler policy. Short version: yes.
Questions
Why build any of this?
Because people are starting to ask an assistant for a marketing agency instead of opening ten tabs. If the assistant can read our services and check our calendar without a human in the loop, we're in that conversation. If it can't, we're not.
Is it free to use?
Yes. No key, no signup, no auth. There are rate limits to stop a runaway loop, and that's it.
Can you build this for us?
That's the point of the page. If you're reading it as a human and wondering whether your business should have one: yes, probably, and we build them. Get in touch.
Want one of these?
We build agent-ready sites for Kuwait businesses — MCP servers, markdown mirrors, structured data, the lot.
Talk to us