For documentation platforms and developer tools

Your docs, exactly right, inside the IDE.MCP tools instead of a scrape.

Developer agents answer questions about your product all day, and today they do it by fetching a rendered docs page and inferring the rest. ADN serves that traffic as MCP tools and code-fenced Markdown, so the answer in the editor is the answer in your documentation — and the crawl behind it stops landing on your app.

  • MCP
  • DocLang (.dclg)
  • llms.txt
  • OpenAPI 3.1
  • Content negotiation

Sound familiar?

Where it hurts today

  • 01

    Agents answer from last year's snapshot

    A model trained on an old crawl, or an agent scraping a rendered page, will confidently give your users an API that you removed two releases ago.

  • 02

    Your search UI is not an API

    Every IDE agent re-derives your structure from HTML — navigation, code blocks, version selectors — and each one gets it slightly wrong in a different way.

  • 03

    Indexing sweeps look like a traffic spike

    Docs platforms are exactly the shape crawlers like: thousands of interlinked pages, all dynamic, all hitting the same database.

What you get

What ADN does about it

MCP tools for IDE agents

Cursor, Claude Code and Copilot call your docs as tools — search, fetch, list — rather than scraping and guessing at structure.

Code-fenced Markdown

Examples survive the trip. No navigation chrome, no cookie banner, no markup noise burning context before the answer starts.

Priority by intent

A developer agent with a human waiting is latency-critical and is never queued behind a bulk training crawl. Both are served; they are not served the same way.

One corpus instead of a sweep

Training crawlers take a bulk export and a delta feed, so re-indexing your docs stops meaning thousands of origin requests.

Two ways in

Whatever is in front of your origin

The pitch changes with your infrastructure, not the product. On Cloudflare we install alongside your zone plan; everywhere else we are the whole agent layer.

Path 1 — Cloudflare add-on

Already on Cloudflare Pro

Keep Pro, Markdown for Agents and SBFM exactly as they are. Pro answers every bot in one format; ADN adds MCP for IDE agents, per-class priority and origin offload on top.

Pro $25 + Growth $149

Path 2 — primary stack

On Vercel, Netlify or bare origin

No Cloudflare bill to add to — ADN is the entire agent layer, installed as npm middleware with a restrictive matcher so it only sees the paths you point at it.

Growth — $149/mo

Where you land

Growth$149/mo

Hosted proxy or edge connector

3 domains, 750k requests, 5k renders, per-class policy engine, bulk corpus export.

When you do not need us

If all you want is a hosted endpoint that turns your site into MCP, that is a commodity — WebToMCP sells it for $9 a month — and we would rather point you at one than pretend otherwise. ADN earns its price when you also want per-class routing, bulk export, offload and traffic-by-class analytics.

Free efficiency scan

See what an IDE agent gets from your docs today

The scan fetches a docs page the way a machine visitor does and shows the token cost of the markup around your content.

Questions

The ones that come up first

We already have Markdown for Agents on Cloudflare Pro. What does ADN add?
Pro converts pages to Markdown and gives every machine visitor the same answer. ADN routes by what the visitor is trying to do: MCP tools for an IDE agent, a bulk corpus for a training crawler, a fast page for a live fetch. Keep Pro — we install alongside it.
Do we have to rewrite our documentation?
No. The content pipeline reads the docs you already publish — sitemap, pages, JSON-LD — and derives the Markdown, the corpus and the MCP tool surface from them.
What happens to our own search and analytics?
Untouched for humans on edge-connector paths. Every failure path fails open to your origin, so a person always reaches your site exactly as before; only classified machine traffic is routed.

See what an IDE agent gets from your docs today

The scan fetches a docs page the way a machine visitor does and shows the token cost of the markup around your content.

Not quite you?