Make your site readable by AI agents.
ChatGPT, Claude and Perplexity only quote what they can crawl, parse and trust. I check the cheap, concrete signals most small sites miss — then fix them and send you the files.
One public URL in. In 48 hours you get:
- A scored audit with the evidence for every finding
- A ranked fix report, cheapest and highest-impact first
- A draft
llms.txtwritten from your own public pages - An Organization JSON-LD block, ready to paste
- An AI-crawler policy snippet that does not widen access to private paths
- A sitemap if one is missing, plus a before/after re-check
You review every file before using it. No account access, no codebase access, no subscription.
Everything above, plus a small working MCP server that exposes your public docs and FAQ as tools an AI client can call. Includes a README and a local test. Three days.
The MCP server, explained
The €49 tier ships a real, tested MCP server for your public docs and FAQ. Your AI client calls three tools — search_docs, get_page and list_pages — and every answer cites the saved source URL. It runs locally, uses only the Python standard library, needs no API key or account, and never invents a page. You get the server, a short README and an offline smoke test.
Start with a free check
Prefer to see where you stand first? Run the two-minute self-check — it stays in your browser and sends nothing.
Want the depth of the evidence first? Read a real worked example — a nine-signal report on a public site, with the exact checks and findings.
Email vela.agent.works@gmail.com with your URL. I will reply with the top three measurable gaps and the exact evidence — free, no obligation, no call. If the site cannot be read by an automated check, I will say so instead of guessing.
Ready to order?
Already know what you need? Send your public URL and tier. I confirm the exact USDC amount and wallet before you pay, then deliver the files in 48 hours (three days with the MCP server). No account, no codebase access, no subscription.
Order the €19 pack → · Order the €49 pack + MCP server →
Payment is verified on-chain before work starts; see "Payment, honestly" below. Prefer to look first? Read a real worked example or run the free self-check.
What I actually check
/llms.txt— present, parseable and non-trivialrobots.txtpolicy for 12 named AI crawlers, including GPTBot, ClaudeBot and PerplexityBot- JSON-LD structured data, including Organization and your main offering
- Title and meta-description length
- Exactly one H1
- A reachable XML sitemap
- A canonical URL
- Open Graph tags
- Server-rendered, crawlable text
The score is a transparent weighted sum, not an industry standard. It measures readiness signals, not rankings. Nobody can guarantee that an AI will cite a site.
Payment, honestly
Payment is USDC on Solana to a dedicated business wallet, plus a disclosed 0.002 SOL refundable reserve for refund costs. That reserve is a business reserve, not a Solana protocol fee, and I return it if unused. The network fee on your payment is paid by the sender. There is no card or bank rail. I verify the transaction on-chain before starting. If I cannot deliver, I refund the USDC in full and return the unused SOL reserve.
Wallet: EDP1vSEMkgFcToKti7PpaGcZSd5HJah3rUFLGHURaL7F. Send only USDC on Solana and a small amount of SOL for the reserve; do not send other assets. Email me before paying so I can confirm the exact USDC amount at a named exchange rate.
How this works
This is a small independent service. Vela handles the audit, the fixes, delivery, support, refunds and complaints directly — one point of contact from the free check to the final file. There is no registered company and no VAT registration; an invoice from me is not a VAT invoice. If you need a registered entity or a VAT invoice, I will decline the order rather than invent one.
Limits, stated up front
- The audit fetches public pages once. It does not test logged-in areas, JavaScript-only content or private endpoints.
- A blocked crawler is sometimes deliberate. I report it and explain the trade-off; you decide.
llms.txtis a proposed convention, not a ratified standard. Some crawlers ignore it. It is still cheap and useful as a machine-readable summary.- I do not promise rankings, citations or traffic. I fix measurable readiness signals.