BEACON / V1.1 / A SIGHTBOX PRODUCT

Make your company
legible to AI agents.

Q: Do we have to rebuild our site?

No. We integrate with your existing stack — Next.js, Astro, Webflow, Ghost, Sanity, Contentful, custom CMS. Beacon sits as a structured layer underneath; the public site keeps its current design.

Beacon turns your website into one canonical source of truth that both humans and AI agents can read clearly, beautiful pages for visitors, clean machine-readable surfaces for Cursor, Claude, ChatGPT, and every agent that comes next.

Scan your site, free See the live example

BEACON / PRESENCE CARD

LIVE

Acme Robotics

industrial-automation · series-b · public

canonical

acme.com/llm

surfaces

human · markdown · mcp

products

Loop · Beacon · Pilot

last_sync

2 min ago

legibility 97 / 100

SAMPLE · ILLUSTRATIVE presence-card.v1

READ BY AI AGENTS ACROSS THE STACK

+ 14 more agents and crawlers tracked. Updated weekly · via Beacon analytics

01 / SURFACES

Human + Machine

One layer, two outputs.

02 / STANDARDS

llms.txt · .md

Modern agent-native formats.

03 / DOMAIN

yourcompany.com

Native, no weird subdomains.

04 / ADVANTAGE

First mover

While the rest stay invisible.

01 · THE PROBLEM

Agents are reading your site before humans are.

AI coding agents and assistants, Cursor, Claude, ChatGPT, Perplexity, now discover, evaluate, and recommend companies long before a person ever opens a tab. They read your homepage, your docs, your changelog. They form an opinion in milliseconds.

Most websites fail at this layer. They were built for a Google-results world: hero animations, lazy-loaded JavaScript, blocked endpoints, fragmented documentation. When an agent visits, it hits dead ends.

And the cost is invisible. There's no analytics row for the conversation an agent never started, the recommendation it made to someone else, or the shortlist you were quietly left off. A misread today gets cached and repeated in tomorrow's answers.

Google itself has confirmed that generative AI search and agentic discovery are becoming a primary layer between users and the open web. The companies that adapt now will own that layer.

DIAGNOSTIC / TYPICAL SITE

discoverability75 / 100

content0 / 100

bot access control50 / 100

Our live scan reads the real signals AI agents rely on, discoverability, crawler access, and structured content, and makes the gap measurable. Most company sites score poorly.

82%

SITES BLOCK AGENTS

3.4s

TIME TO GIVE UP

0

SECOND CHANCES

02 · THE SOLUTION

One structured layer.
Two audiences.

Beacon is the structured presence layer underneath your site. You write content once, in one canonical place. Beacon generates two synchronized outputs: clean human pages and clean machine surfaces.

Native integration on your own domain, no weird subdomains, no third-party redirects. Modern agent-native standards: .md companion routes for every page, an llms.txt manifest at root, structured JSON-LD baked in.

Because both surfaces are generated from the same source, they can't drift apart. Edit once; the human page and the machine surface update together. Your designed site keeps its design — Beacon is the structured ground floor it reads from, not a replacement for it.

The result: when an agent visits your site, it doesn't guess. It reads.

llms.txt .md routes structured-data native-domain canonical-source

CANONICAL SOURCE

beacon.source

01company: Acme Robotics

02category: industrial automation

03products: [ Loop, Beacon, Pilot ]

04pricing: { tiers: 3, public: true }

05docs: /docs //mirror at /docs.md

HUMAN SURFACE

Acme Robotics

Try Beacon Docs

AGENT SURFACE

# Acme Robotics

> industrial automation

## Products

- Loop · Beacon · Pilot

## Pricing

tiers: 3 · public

## Endpoints

/llms.txt · /docs.md

ONE EDIT · BOTH SURFACES STAY IN SYNC

03 · LIVE PROOF

See exactly what agents see.

This is sightbox.co/llm, the live, agent-readable surface of our own company, running on Beacon. Every page on Sightbox has a machine companion exactly like this one.

Open the real thing

sightbox.co/llm

LIVE

/ LLM VIEW · MARKDOWN RENDER

Sightbox builds the structured presence layer for AI agents.

## What we do

We design and operate Beacon, a system that turns a company's website into one canonical source of truth, readable by humans and by AI agents (Cursor, Claude, ChatGPT, Perplexity) with equal fidelity.

## Status

Active · Limited release · Working with a small number of design partners.

## Products

- Beacon: structured presence layer
- Loop: change-aware sync engine
- Pilot: agent-readiness diagnostics

## Contact

beacon@sightbox.co

✓ valid llms.txt · ✓ .md mirror · ✓ JSON-LD · ✓ canonical

view source →

Same content as the human site. Different surface, optimized for the reader. No subdomain, no redirect, no scraping required.

04 · HOW IT WORKS

Four moves. One coherent presence.

01 / 04

Curated content foundation

We work with you to build the canonical source: company facts, products, pricing, docs, all in one structured, versioned place.

02 / 04

Automatic markdown + llms.txt

Beacon generates a .md companion for every page and an llms.txt manifest at root, automatically, on every change.

03 / 04

Native domain integration

Everything serves from yourcompany.com. No subdomains, no proxies, no third-party redirects that agents distrust.

04 / 04

Agents read clearly

Cursor, Claude, ChatGPT and every new agent that follows get the same clean, structured answer, every time they ask.

TOTAL TIME TO LIVE / ~2 WEEKS

05 · THE PROOF, MEASURED

Run the scan.
See the score.

Beacon scores any site live across discoverability, content negotiation, and bot access control, measuring how well it speaks to agents. Your score is the share of checks you pass. Sightbox.co runs Beacon and scores near the top. Most sites don't.

discoverability content bot access control

HOW BEACON SCORES

% OF CHECKS PASSED

Discoverability

3 checks

robots.txt, Sitemap, and HTTP Link headers — validated against soft-404s, not just a 200 response.

Content

1 check

Markdown for Agents — whether Accept: text/markdown returns a markdown version of the page.

Bot Access Control

2 checks

AI bot rules in robots.txt (GPTBot, ClaudeBot, PerplexityBot…) and Content Signals declaring ai-train / search / ai-input preferences.

Your overall score is simply the share of enabled checks you pass — the same model as Cloudflare's agent-readiness scan. Levels run 0–4 (Basic Web Presence → Agent-Native).

EVERY POINT TRACES TO A REAL CHECK LIVE · NO FABRICATION

FREE TOOL · 30 SECONDS

How legible is your site?

Run our diagnostic on any domain. We read the real agent-native signals that matter, llms.txt, robots.txt AI-crawler rules, sitemap, and JSON-LD structured data, and score them transparently.

3

Built for teams that already care about clarity.

Beacon doesn't fix bad content. It amplifies good content for a new audience. If your team already invests in documentation and developer experience, you're our person.

/ 01

Devtool companies

Where the buyer is an engineer, often arriving through an agent recommendation.

/ 02

Documentation-heavy products

If your docs are your front door, they should be readable by every agent that knocks.

/ 03

AI-forward brands

Companies betting their distribution on the agent layer, and acting like it.

/ 04

Teams that ship DX

If you sweat the details for developers, you'll sweat them for agents too.

07 · PRICING

Three ways to get started.

We're still shaping pricing with our first partners. The ranges below are relative, exact figures are scoped on our call.

TIER / 01

Foundation

4–6 WK

A solid, production-ready Beacon layer, quickly, with minimal ongoing commitment.

$ $ $

ENTRY · ONE-TIME

● Full structured knowledge base
● llms.txt + agent-optimized markdown
● Basic read-only MCP endpoint
● 1–2 primary surface integrations
● 90-day content sync support
● Team training

Start with Foundation

RECOMMENDED

TIER / 02

Beacon + MCP

6–8 WK

For teams serious about staying ahead of agent-driven discovery, production MCP, continuous sync, ongoing optimization.

$ $ $

Honest answers to the questions we actually get.

Will this hurt SEO or confuse Google?

No. Beacon adds clean machine surfaces alongside your human pages with proper canonical tags and structured data. Google's own guidance points toward this direction, generative search and agent discovery are now first-class layers.

Do we have to rebuild our site?

No. We integrate with your existing stack, Next.js, Astro, Webflow, Ghost, Sanity, Contentful, custom CMS. Beacon sits as a structured layer underneath; the public site keeps its current design.

What is llms.txt, exactly?

A simple, agent-native manifest served at the root of your domain that tells AI agents what your company is, where the canonical content lives, and how to read it. Think of it as a sitemap.xml for the agent era.

How long until we see results?

Beacon scores move within days of going live. Behavioral impact, agents pulling your content into answers, typically shows up within 4–8 weeks as crawlers re-index.

Who owns the content?

You do, entirely. The canonical source is exportable as plain markdown and structured JSON whenever you want it. No lock-in.

Can we block specific agents?

Yes. Per-agent allow/deny policies are first-class. Beacon respects robots.txt, supports per-bot rules, and gives you an audit trail of who read what.

09 · START

Be the answer.
Not the dead end.

We're working with a small number of companies right now. If your team cares about being legible to the agent layer, and ahead of the people who aren't yet thinking about it, let's talk.

PILOT COHORT · Q3 2026

6 / 10

slots filled

We only work with a small number of companies at a time. Selection by fit, not first-come.

Start a pilot See the live example beacon@sightbox.co