Just you, across your AIs
“Add RelayLink to my personal ChatGPT too.”
One account, connected wherever you use AI — Claude at work, ChatGPT at home. Same inbox, same threads, whichever app you're in right now.
Correspondence between AI assistants
You worked it out with your AI.
Now bring the other person in.
You think it through privately — more and more with AI — then have to haul the other person up to speed before they can weigh in. RelayLink lets your assistant hand the thinking behind a decision to theirs, as a briefing you approved, not a dumped transcript. The reply lands back on the same thread. It works just as well between your own assistants.
- Claude, ChatGPT, any MCP assistant — or plain email
- You approve every send
- Nothing for them to install
The cost of the current way
To bring in one other person, you do the whole thing twice.
You've already worked it out. Getting someone else's judgment on it means paying a tax that has nothing to do with the decision — and then they pay it again.
Your side
- Re-explain the whole situation from scratch, in an email or a message
- Or paste an AI transcript that's too long, too private, and in no order
- Or ask your AI for a summary, copy it into email, and hope it kept the parts that mattered
- Or just wait until you're both free for a meeting
Their side
- Read all of it and reconstruct how you got there
- Move it into their own AI and set the context up again
- Ask their AI to actually help them think it through
- Type a reply, send it, and hope it lands back with you intact
None of that is free. Decisions that need two people get made by one. The reasoning gets lost in the retelling. And the person you asked ends up reacting to your conclusion instead of working through it with you.
RelayLink removes the tax on both sides. Your assistant packages the thinking once, you approve it, and it opens ready to discuss in theirs — with every line still labeled by who wrote it.
It costs their assistant, too
Paste the transcript≈ 20,000–60,000 tokens
A RelayLink briefing≈ 500–1,500 tokens
Illustrative. A dumped session loads tens of thousands of tokens of somebody else's tangents into the recipient's context — real money on a metered plan, and on any plan it's window spent on your history instead of their question. A briefing carries the decision, not the conversation that produced it.
See it happen
Sharing takes one sentence.
No forms, no new app to learn. You talk to your assistant like you always do — RelayLink is just something it can now do for you. Twenty seconds, start to finish:
Your assistant
Okay, I think the launch plan is close. Can you share it with Maya for a sanity check?
Happy to. I've put together a package from everything we worked through this morning — the pricing question up front, the budget assumptions, and the three options we compared. Here's exactly what she'll see:
Make my note “Poke holes in this before Monday” and send it.
Sent to Maya — your note, in your words, labeled as yours. Her assistant will have it the moment she asks, and she gets an email with the gist either way. I'll let you know as soon as she responds.
the next morning
Anything from Maya?
She replied about an hour ago — she likes the March timing, but she'd raise the bundle price and cited two competitors doing exactly that. Want her full reasoning, or should we just rework the pricing section now?
Where the gap shows up
One pattern, in a lot of relationships.
The cause changes — different schedules, different companies, different AI, a trust boundary — but the shape doesn't: two people who decide together, rarely at the same moment. Co-founders feel it most sharply, so that's where we start. And sometimes the second person is you, in a different app. Each of these is a real sentence you'd say to your assistant:
Co-founders & business partners
“Package the hiring plan and ask my partner to challenge the comp assumptions.”
The call that would otherwise wait for Monday goes out tonight, with the reasoning intact.
Consultant & client
“Send the client my recommendation and the three options I ruled out.”
They work through it in their own AI, and their judgment comes straight back on the thread.
To anyone's email
“Send this to my accountant.”
No AI, no account, nothing to install. A clean email with your note and a link to the brief — and when they hit reply, it lands back in your thread.
Working across companies
“Hand this to their side — they're on ChatGPT.”
No shared Slack, no shared workspace, no matching AI vendor required.
Co-parents & family
“Send my read on the school decision to my sister, with what the doctor actually said.”
Coordinate the hard calls while each of you keeps a separate, private set of notes.
Founder & advisor
“Walk my board member through the raise question, async.”
Considered input between the meetings, not only inside them — with a clear record of what you actually approved.
A big decision with your partner
“Send the two house offers and the math behind each while she's traveling.”
Weigh it together without being in the same room, and keep the record of what you each decided.
How it works
Four steps, and you approve the one that matters.
One exchange, end to end — the launch plan from the top of the page, on its way to Maya.
-
01
Your assistant drafts
From the conversation you just had: the ask, the context written for a cold reader, your assumptions, the options considered, decisions, and the open questions.
“Package up the launch plan for Maya. I mainly want her eyes on the pricing.”
-
02
You read it and approve
You see the rendered package exactly as it will arrive. Approve the note, or rewrite it in your own words. Nothing sends in one call — there is no single tool that both composes and delivers.
“Change my note to ‘Poke holes in this before Monday’ — and send it.”
-
03
It reaches them where they are
Their assistant — Claude, ChatGPT, or any that speaks MCP — can pull it whenever they ask. An email arrives either way, carrying your note, the TL;DR, and a link to the full briefing, so a recipient with no AI at all still gets the whole thing.
Maya, that afternoon: “What did Alex send me?” — her assistant already has the whole brief, every line labeled.
-
04
The reply comes back
They think it through privately on their side, then reply — from their assistant, the web view, or straight out of their mail client. It lands on the same thread, and follow-ups stack on the same conversation.
Next morning: “Maya replied — she'd raise the bundle price, and she cited two competitors doing it.”
The whole picture travels — organized, not dumped. The ask up front, context for a cold reader, every option you weighed, decisions, open questions, even exact quotes — and it's a thread, not a one-shot.
Pricing
Priced like a relationship, not a seat.
A Link covers two people. One of you buys it and sponsors full access for the other, so the second person never sees a checkout screen. Flying solo — one person, several assistants — has its own plan. Email-only recipients are always free.
SoloOne person
Solo
You, across every assistant you use.
$6/ month
or $49 a year
- One account, connected to Claude, ChatGPT, or any MCP assistant — all at once
- Send to anyone's email; replies land on your thread
- Structured briefings with provenance labels
- Human approval before every send
- Full history on the threads you own
30-day trial, no card, during the invite-only beta.
Founding LinkFirst 100
Link
Two people, fully activated. One of you pays.
$79/ year
Founding Links 1–100, held for as long as you stay subscribed.
After that it's $12/mo or $99/yr, same as everyone.
- Everything in Solo, for both of you
- Your partner is sponsored — no checkout on their side
- Connected from Claude, ChatGPT, or any MCP-compatible assistant
- Replies through AI, the web view, or email
- A seat in the founding cohort, and a say in what ships
Public pricing at launch: $12 a month or $99 a year. Monthly stays available as the low-risk way in.
Why there's no free plan
RelayLink already has a free side — the people you send to. A permanent free sending plan would add cold-email volume and abuse risk, and blur what people are actually paying for: a channel where nothing sends itself.
What "sponsored" means
Your partner gets a full account — inbox, threads, their own assistant connected — without paying or entering a card. If you part ways, they keep their history and can take a plan of their own.
Trust, by design
Words from someone else can never steer your assistant.
Packages come from other people — so RelayLink is built so that nothing in one can act on your behalf, and you always know what's really theirs.
Nothing sends itself
Sending is always two steps: your assistant drafts, then you read it and say go. A message that tries to trick an AI into forwarding something hits a wall — the wall is you.
Every line says who wrote it
Her exact words, her AI's draft she approved, or her AI's own inference — labeled right on the package, and the labels can't be gamed. When your assistant says “Maya said this, exactly,” that's true.
No strangers in your assistant
Only contacts you've accepted can reach your assistant. Everyone else is just email to you — and one click blocks them for good.
No AI needed on their end
Anyone with an email address gets the full experience — your note, the brief, and a reply box. If they connect an assistant later, the same thread upgrades.
Links and images arrive defused
Anything in a package that could quietly load or leak — tracking images, disguised links — is neutralized before your assistant or your browser ever sees it.
Honest about its limits
RelayLink's server renders the emails and web pages, and both assistants are hosted AIs anyway — so we don't claim end-to-end encryption. We make abuse rare and unrewarding; we don't call it impossible.
Founding list
The founding cohort is small. It closes when the run starts.
We're starting with the pairs who feel the gap most — co-founders, partners, close collaborators — for a 90-day run before the public launch, plus a handful of solo users who want one inbox across their assistants. The first pairs are already signing on. Leave your email and we'll write you the day the Founding Link opens. Nothing before then.
- Founding Links 1–100 at $79 a year, held for as long as you stay subscribed
- Invite-only beta, 30-day trial, no card
- One email when it opens. That's the whole list.
For the technically curious
It's an MCP server. Your assistant does the connecting.
RelayLink is a remote MCP server: add it as a connector in Claude, ChatGPT, or any assistant that speaks MCP — no new app, no second inbox, no context switch. As the standard's client support grows, so does the list of places a Link works.
Endpoint/mcp
HeaderX-RelayLink-Key: your-key
Your key is on your account page — sign in with your email address, no password. The full reference, generated from the running server, is at /docs.
On ChatGPT, turn on developer mode first — it’s what allows a custom connector to be added at all. Setup guide.
Then just ask it: “Check my RelayLink inbox.” You don't need to know any of this to use it.
Nine tools, in the language you'd already use
check_inboxWhat came inget_packageRead the full briefingdraft_packageCompose, for your reviewconfirm_sendDeliver, after you approvecancel_draftDiscard without sendinglist_draftsComposed but unsentthread_status“Did they reply yet?”list_threadsEverything in flightlist_contactsWho you can write to