Free · 5 minutes · no code
Is your site invisible to AI?
Free 5-minute score. See what ChatGPT, Claude, and Perplexity can't find on your site — and ship the fix today.
For agent builders
Skip the crawl. Query Lattis.
Every indexed site is a live MCP endpoint — owner-verifiable, freshness-stamped, ready for Claude, Cursor, Claude Code, or any MCP client. One line of config. No API key.
claude mcp add lattis --transport http https://mcp.lattis.dev/mcpFrom URL to agent-ready in five minutes
No engineers. No code changes. No MCP to babysit.
Your website
Paste any public URL
Lattis reads & grades
AI agents find you
Claude, ChatGPT, Cursor, and thousands more
What the score actually measures
A 0–100 grade for how well agents can answer questions about you
Not a vibe check. A reproducible rubric we run on every site, including our own — see Lattis's own score at app.lattis.dev/sites/lattis.dev/report.
01
We generate ~15 buyer questions
A category-aware model writes the questions actual buyers in your space ask before they buy. Pricing, security, integrations, edge cases, the gotchas.
02
A judge model grades each one
Every question runs against your indexed content. Each gets scored answered, partial, or unanswered — with a verbatim quote from your site as evidence.
03
Each miss gets a diagnosis
Missing page, vague claim, contradictory pages, collapsed content (trapped in a tab agents can't expand), or genuinely absent. Plus a per-question fix.
04
Score is the weighted total
0–100. High-importance buyer questions count more. Rolls up into a single number you can track over time, with a label from Invisible to Excellent.
The diagnosis
The 5 ways your site confuses agents
SEO grades you against keywords. Lattis grades you against questions buyers actually ask, and every miss gets tagged with one of these — a named reason, per question, with a fix hint you can hand to a writer or an engineer.
No dedicated page exists for the topic. Buyers ask; your site has nothing to serve.
A buyer asks for rate limits; you have API docs but no rate-limit page.
The right page exists, but the body is hidden behind JavaScript that never rendered.
Your integrations list is inside an accordion that crawlers (and agents) can't see.
The topic is mentioned, but the specific number or fact a buyer needs is missing.
“Generous free tier” with no limits. “Competitive rates” with no percentages.
Two or more pages give different answers. Agents pick one at random.
README says Node 18+, install docs say Node 20+. Which one is right?
The topic isn't anywhere on the site. Not a page gap — a story gap.
No mention of SOC 2, chargebacks, or whatever an enterprise buyer just asked.
Stop maintaining your own MCP
Self-built MCPs rot. Lattis doesn't.
The day after you ship a feature, every hand-rolled agent interface is already wrong. Lattis is recrawled on your schedule — no sprint, no maintenance, no stale answers.
Someone on your team has to notice, write a ticket, ship a refactor, redeploy. Meanwhile every agent is serving 47-day-old answers.
One recrawl, and every MCP response carries a last_crawled field agents can trust. No babysitting, no drift.
Claim your domain
One DNS record unlocks the rest
You get a score for free. Verify ownership with a single TXT record to expand crawls, push re-crawls on demand, and bring your team in. No code, no plugins.
Crawl 2,000 pages instead of 500
Free tier crawls 500 pages. Verifying ownership via DNS bumps you to 2,000 — enough to reach the security pages, changelog, deep docs, and edge cases buyers actually ask about.
On-demand re-crawls
Ship a feature, hit re-crawl, your score updates within minutes instead of waiting for the next natural refresh.
Team workspaces
Manage multiple verified domains under one organization. Invite teammates, share reports, track score over time.
Remove stale evaluations
If a question stops applying or the answer has moved, drop it from your report and re-grade against your real surface area.
Two sides stuck. One layer between them.
Agents and websites keep missing each other
The pain shows up in different places, but it's the same gap — there's no trusted, fresh interface between the two.
Every new site is another integration you hand-roll.
- −Hand-wiring MCP servers for every site you touch
- −Falling back to WebFetch, burning tokens parsing HTML
- −Installing a vendor MCP that's already stale by the next release
The moment you ship a feature, every agent forgets.
- −SEO, meta tags, and PageRank mean nothing to an agent
- −Shipping your own MCP takes a sprint you don't have
- −The day after you ship, your agent interface is out of date
Common questions
Do I need to change anything on my website to use Lattis?
No. Lattis works with any public website as-is. Paste your URL — we crawl, chunk, embed, and serve your content to AI agents. No code, no meta tags, no SDK, no DNS changes required to get a score.
How is Lattis different from traditional SEO?
SEO optimizes for human search engines using rankings and meta tags. AI agents don't use rankings — they read content directly. Lattis grades how well your site answers the questions buyers actually ask and serves the content to agents through a hosted MCP endpoint. SEO for AI.
Which AI tools and platforms work with Lattis?
Claude, ChatGPT, Cursor, Claude Code, Windsurf, and any client that speaks MCP, REST, or stdio. See lattis.dev/builders for one-line install snippets per client.
Can I use Lattis with multiple websites?
Yes. The same hosted MCP endpoint covers every indexed site, and you can submit any number of public domains for indexing. Verify each one via DNS to manage them under a team workspace.
Where's the documentation?
Full integration guide at lattis.dev/docs — hosted MCP server, per-site endpoints, REST API, CLI, and install snippets for every supported client.
Stop meeting in the middle the hard way.
Whether you build agents or run a site, the layer between them should already exist. That's Lattis.
Free for public websites. No credit card. No code changes.