How to ask RelayLink for help

Open a support request from the public form, your account page, or your assistant. The printed reference is a convenience. The Reply-To token is what lets you add to the thread by email. Acknowledgements echo nothing you typed.

3 min read

Something is wrong โ€” a connector that will not connect, a package that never arrived, a question the docs do not answer. Start at /support. /help and /contact redirect there. The footer on every marketing page links it. Signed in, the same form is one tap from every account page.

hello@relaylink.ai stays on the legal pages and in the acknowledgement, because somebody whose problem is that RelayLink is not working needs a path that does not go through RelayLink.

Three ways in

The public form. No account required. No cookie either โ€” an anonymous marketing page that set a cookie on every visitor would stop being cacheable. What bounds it: a honeypot field, a per-IP rate limit, and a per-address daily cap. The address you type is how we write back, not proof that the account is yours. An anonymous request carries no account facts, whatever address is typed.

Signed in at /account/support. The session already knows who you are.

From your assistant. open_support_request, support_status and reply_to_support are three of the thirty-eight tools. A support request is written to us, is not a package, and needs your own word. There is no tool for the assistant to answer as RelayLink: an answer under the company's name should cost a person typing it.

The reference is not a key

You will see a short reference like RL-7Q4M2X in mail, in a link, and on a page. It is so you can find the thread. It authorises nothing. Asking for a request that is not yours returns the same answer as asking for one that does not exist.

What lets you add to the thread by email is the Reply-To address on the mail we send you. That token appears only there. A matching signature on inbound mail proves the message reached us; it does not prove who typed it. Forwarding and aliases are ordinary, so a mismatched From is kept and flagged, not thrown away, and never treated as "the account holder said this."

Our reply always goes to the address stored on the request, never to an inbound From. Nothing arriving by email may authorise account recovery, an ownership change, a key action, or the disclosure of any private account fact.

What we keep, and what we do not

If the browser sends diagnostics, the server keeps a small allowlisted set โ€” not a dump, and not identifiers from paths like /account/packages/{id}. The acknowledgement you receive echoes none of that.

You get the reference and a way to add more: reply to the mail, open the thread in your account, or tell your assistant to reply. If mail to you cannot be delivered, the desk sees that. "The provider took it" is not the same as "they got it."

For connector failures, troubleshooting MCP connection problems often settles it before a ticket. For everything else, open a request or ask your assistant once you are connected.

Frequently asked questions

Does the reference number let someone else read my request?
No. A reference like RL-7Q4M2X is printed in mail and on pages so you can find the thread. Reads are scoped to the owning account. A request that is not yours answers exactly like one that does not exist. What authorises an emailed reply is a Reply-To token that never appears in a query string.
Will the acknowledgement email quote what I wrote?
No. It goes to an address nobody verified, so it carries the reference, that we have it, and the fallback address — never the submitted message or any account facts. A stranger typing your address into a public form must not become a way to make us mail you your own words.
Can my assistant send a support reply under the company's name?
No. You can open a request, check status and add to a thread from your assistant. Answers from RelayLink are typed by a person. That is deliberate.