Your AI, your people, one thread

Conversations for
the AI era.

Keep your ideas, decisions, and conversations together. Bring the right context to someone else, get their input, and continue the work with your AI.

  • Any AI you connect
  • Anyone, by email
  • You approve every send

$164saved a year

Saved, not spent. Sending briefings instead of pasted transcripts costs about that much less in tokens on typical use, because an assistant re-reads what you send on every turn.

See how that is worked out

Where the idea stands

  • Your idea Neighbors lend each other tools instead of buying them.
  • Jim’s input Try it with a few neighbors before building anything.
  • Still open What happens when a borrowed tool comes back broken?

Example. It is the first story below.

How it works

What are you working on?

See how an idea, a decision, or a project moves forward when you bring someone into the thinking.

“Help me think this through.”

  1. Think it through

    YouWhat if neighbors could borrow tools from each other instead of buying things they’ll barely use?

    Your AIWorth pulling apart. Who is it for — a street, a building, a whole neighborhood? One option is a small app for a defined group. What we cannot answer yet is whether people would actually lend.

    You explore who it is for, possible approaches, and what you do not know yet. The app is one idea, not a plan.

  2. Save the useful parts

    YouSave the idea, the approaches we considered, and the questions we still have.

    Saved to your notebook

    • GoalNeighbors lend each other the tools they rarely need.
    • OptionA small app for one neighborhood.
    • OpenWould people actually lend? What happens when something comes back broken?

    An explicit save of the parts you want back. Nothing is copied in on its own.

  3. Review and share

    YouShare this with Jim. Include the approach we considered and ask what I might be missing.

    Ready to send — read it first

    ToJim

    Your note“What am I missing here?”

    IncludedThe goal, the app option, both open questions

    Approve and sendSent

    Later, with Jim

    JimI like it. But people might worry about getting a tool back broken. Could we try it with a few neighbors before building an app?

    Jim works through it with his own AI and approves that reply himself.

  4. Continue with their input

    YouLet’s work through Jim’s feedback and save an updated proposal.

    Your AIPulling up the saved idea and his reply. His worry is broken tools, and his suggestion shrinks the first step: a few neighbors, before any app.

    Where it stands

    • The ideaNeighbors lend each other tools.
    • New perspectivePeople may hesitate if a tool could come back broken. Jim’s reply
    • Proposed next stepTry it with a few neighbors before building anything. proposal
    • Still openWhat happens when a borrowed tool comes back broken?

    The first version is kept. This one stays a proposal until Jim accepts it.

Their perspective becomes part of your next conversation.

“Could this room work for both of us?”

  1. Think it through

    YouHelp me turn the spare room into a home office. Budget: $1,500. No major renovations.

    Your AITwo layouts worth weighing. Desk against the window with storage on the long wall, or desk in the corner, which leaves the middle of the room clear. Both depend on what the room actually measures.

    Layouts to consider, not a shopping list.

  2. Save the useful parts

    YouSave the budget, the layouts, and what we still need to measure.

    Saved to your notebook

    • GoalA home office in the spare room.
    • Constraints$1,500. No major renovations.
    • OptionsDesk at the window, or desk in the corner.
    • OpenMeasure the usable wall and the door swing.
  3. Review and share

    YouShare the options with Pam and ask what they think.

    Ready to send — read it first

    ToPam

    Your note“Either of these work for you?”

    IncludedThe budget, both layouts, what we still have to measure

    Approve and sendSent

    Pam replies by email — no AI, nothing to install

    Pam, by emailI’d love a better workspace too. But we also need somewhere the kids can play. Could the room do both?

  4. Continue with their input

    YouLet’s explore Pam’s suggestion using the budget and layouts we already considered. Save the options we want to discuss.

    Your AIDirections worth looking at: a foldaway desk, storage you both use, and a room that changes through the day. What any of them costs is the next thing to check.

    Where it stands

    • Your original goalA home office.
    • Pam’s inputKeep space for the kids. Pam’s email
    • Direction to exploreOne room that serves both needs. proposal
    • Still openWhich layout gives everyone enough space?

    Budget stays $1,500. Major renovations stay out.

Make room for another perspective without losing your priorities.

“Let’s figure out what your customers need.”

  1. Think it through

    YouHelp me plan Bob’s website. The goal is more qualified inquiries.

    Your AIA homepage that introduces the company, explains what it does, and invites people to get in touch. What we do not know is where their customers currently stall.

  2. Save the useful parts

    YouSave the goal, the proposed page structure, and the questions we need the client to answer.

    Saved to the project

    • GoalMore qualified inquiries.
    • Proposed approachIntroduce, explain the services, invite contact. proposal
    • Ask BobWhere do potential customers usually get stuck?

    The goal and the approach are separate entries, because only one of them is settled.

  3. Review and share

    YouShare the approach and ask where Bob’s customers usually get stuck.

    Ready to send — read it first

    ToBob

    Your note“Before I go further — where do your customers usually stall?”

    IncludedThe goal, the proposed structure, the open question

    Approve and sendSent

    Bob reads the shared background and replies

    BobPeople usually do not know which service they need. We spend most of the first call helping them figure that out.

  4. Continue with their input

    YouHow should that change our approach? Save a revised proposal and what we should ask next.

    Your AIIt moves the first job of the page. Instead of listing services, it helps a visitor work out which one they need and takes them there.

    Where it stands

    • The goalMore qualified inquiries.
    • Bob’s insightCustomers need help choosing a service. Bob’s reply
    • Revised proposalHelp visitors find the right path first. proposal
    • Still openWhich customer questions should guide those paths?

    The first approach is kept in the history, with the reason it changed.

Keep the reason behind the direction everyone is discussing.

Illustrative walkthrough. Your AI’s wording may vary. Nothing here sends a message, connects an account, or calls a live AI.

Features

More of your thinking, put to use.

Eight ways to keep useful work available, involve the right people, and make the next step clearer.

Context-rich conversations

Think it through together.

Share the idea, the background, and the part you need help with. Bring the other person’s perspective into your next AI conversation.

“Share the approaches we considered and ask Jim what we’re missing.”

See a story

Private notes and preferences

Remember what matters to you.

Save the goals, preferences, lessons, and decisions you want your AI to use again. Keep useful thinking ready for your next session.

“Save why we chose this approach and when we should reconsider it.”

How saved work is kept

Cross-AI memory

Pick it up in another AI.

Retrieve your saved work through another compatible assistant. Your notes and handoffs stay with your RelayLink account.

“Pick up the project using my saved notes and latest handoff.”

How it moves across

Search and structured records

Find the decision. And the reason.

Ask for the idea, choice, or unanswered question you remember. Retrieve the relevant record with its source and surrounding context.

“Why did we decide to try it with a few neighbors first?”

See a result

Questions, tasks, and follow-ups

Keep loose ends in sight.

Track what still needs an answer and who you are waiting on. Set a reminder to see when you return to RelayLink.

“Track the question about broken tools. Remind me when I come back after Friday.”

What returning looks like

Selective sharing

Share the useful parts.

Choose the notes and background someone needs. Review the outgoing message while the rest of your private notebook stays on your side.

“Share this proposal and its assumptions. Leave my other notes private.”

What stays private

Email replies and forwarding

Bring email into the work.

Let someone reply by email, or forward a selected message for your AI to work with. Your personal mailbox can stay disconnected.

“Read the client email I forwarded and compare it with our saved plan.”

How people reply

Versions and linked thinking

See how the idea changed.

Keep earlier versions, related explorations, and the lessons behind a new direction. Return to the original reasoning as the idea develops.

“Show the original idea and the changes we made after that feedback.”

Follow a result back

Every feature, on a page of its own

Why not email?

Keep the thinking with the conversation.

Email can carry the exchange, and an AI connected to your inbox may be able to find it. RelayLink adds a place for the notes, decisions, preferences, and open questions you choose to keep alongside that correspondence.

Eighteen seconds, no sound. The same decision sent the usual way, then as one briefing you approve. Jim writes it; Pam answers in the same thread.
  • When you want toReturn to a decision

    An email-based approachFind and read the relevant discussion, or maintain a separate summary.

    What RelayLink addsA recorded choice and rationale, linked to its source.

  • When you want toInclude private working notes

    An email-based approachPut them in a draft, send them to yourself, or keep them elsewhere.

    What RelayLink addsA private notebook available alongside your correspondence.

  • When you want toContinue through another AI

    An email-based approachGive that assistant access to email and any separate notes it needs.

    What RelayLink addsThe same saved records, through a compatible connection.

  • When you want toFollow an unresolved question

    An email-based approachSearch the exchange or maintain a separate follow-up list.

    What RelayLink addsA tracked item tied to the original question.

They can keep using email. You can keep the work together.

The token bill

Send the decision, not the transcript.

When you bring someone in, their AI has to read what you send. A briefing is small on purpose — and it is small without leaving the reasoning behind.

What their assistant has to read

Paste the transcript 30,000 tokens

Typically 20,000 to 60,000 · 9¢ every time it is read

Send a RelayLink briefing 1,500 tokens

Typically 500 to 1,500 · 0.45¢ every time it is read

And it still carries

  • The ask
  • The reasons
  • Options you ruled out
  • What is still open
  • Who wrote every line

What that adds up to in a year

$173

pasting transcripts

$9

sending briefings

$164

the difference — more than a Pro seat costs

Twenty exchanges a month, each re-read about eight times as the conversation goes on, at $3 per million input tokens — a mid-tier rate when this was written; the current ones are published by Anthropic and OpenAI. Counts depend on the model and on how you work; the shape of the difference does not. Whoever’s assistant reads it pays that bill: yours on a hand-off to yourself, theirs on a send.

How this is worked out
  • A pasted transcript 20,000 to 60,000 tokens — roughly 15,000 to 45,000 words. A long working session, not a short exchange. 30,000 is the middle of that.
  • A briefing 500 to 1,500 tokens: the ask, the reasons, the options ruled out, what is still open, and who wrote each line.
  • Read on every turn An assistant re-reads the whole context each time it answers, which is why providers sell prompt caching to make resending it cheaper. Eight turns is a short conversation.
  • Input price $3 per million tokens, a mid-tier rate when this was written. Current rates are published by Anthropic and OpenAI.
  • The multiplication 20 exchanges a month × 12 months × 8 turns × 30,000 tokens × $3 per million = $172.80. The same at 1,500 tokens = $8.64. The difference is $164.16.

Change any assumption and the answer changes; the shape of it does not. Use your own numbers — that is why every one of them is listed.

  • ≈20×

    Smaller to read

    A briefing is the decision. A transcript is the whole afternoon, tangents included.

  • Every time

    Not just once

    Their assistant re-reads the context on each turn. The saving repeats; the paste does too.

  • Nothing

    Private lines along for the ride

    Your session stays yours. What travels is what you approved, labeled by who wrote it.

Cross-AI memory

The AI can change. Your saved thinking stays with you.

Claude at work and ChatGPT at home each end up remembering a different half of you, and neither is built to read the other. RelayLink is the one thing both connect to for the same account.

You choose what is kept, either way. RelayLink does not copy your chat history.

How cross-AI memory works

Set it up in the AIs you use

Handing work from one AI to another

Search and records

“Why did we choose that?” Find it in the work you saved.

Ideas evolve. Keep the choices, rejected options, questions, and reasons organized so your AI can retrieve the part you need. The result names its sources and what is still open.

Keeping a decision log your AI can search

Follow-through

The question is still open. Keep it in sight.

Track a question, a task, or a reply you are waiting for, and keep it connected to the conversation that created it. Reminders appear when you return or when you ask a connected assistant — RelayLink does not email you or start an AI session on its own.

How follow-ups are tracked

Working with people

Bring someone into the thinking. They can reply their way.

Share selected context with a friend, partner, client, or teammate. They can work through it with their own connected AI, or read and reply through email or the browser.

Their own assistant

They connect the assistant they already use, and the shared background is there when they ask for it.

A reply by email

Somebody who has not joined RelayLink reads your note and the briefing, and answers from their own mail. Nothing to install.

A request link

Publish a link for feedback, a proposal review, or a question. The submission arrives in RelayLink, ready to work through.

One person, or a group

Use a shared conversation for a group, or separate conversations when replies should stay private. That choice is fixed when the group is created, and the review step names it before anything is sent. Removing someone later does not retract what has already been delivered.

An email worth bringing over

Forward a selected message to your private RelayLink address and ask your AI to work with it alongside your saved notes. It brings over the text of what you forward — not attachments, and not your mailbox.

Shared with members, or private to you

Project decisions and constraints can be shared with the members who need them. Your notebook is separate, and stays on your side unless you share an entry from it.

Existing RelayLink accounts control who can reach them with accepted contact requests and blocking. A recipient who has not joined can read and reply without connecting anything.

Working with someone who uses a different AI

Connect

One address, into the AI you already use.

Get started creates the account. Connecting is a later step: paste this address into Claude, ChatGPT, or another compatible assistant. The setup guide has every client.

Endpointhttps://relaylink.ai/mcp

HeaderX-RelayLink-Key: your-key

RelayLink connects to compatible AI tools through MCP. Depending on the client, access is authorized with a sign-in or an API key. Claude and ChatGPT are the two documented paths; they are examples, not the whole category. On ChatGPT, turn on developer mode first — it is what lets a custom connector be added at all.

Your key is on your account page once you are signed in. The walkthrough is at /docs.

No account yet? Get started.

Security and privacy

Your thinking. Shared on your terms.

Choose what to save, review what you send, and control which assistants stay connected to your account.

  • TLS with HSTS
  • Encrypted at rest
  • No passwords anywhere
  • Per-account audit trail

Review before sending

Sending uses separate draft and approval steps. Read the outgoing message and its selected context before it goes. Nothing sends itself.

Keep private notes on your side

Share chosen entries or project facts. Sharing does not give the recipient access to the rest of your private notebook, or to the AI conversation it came from.

Control your connections

See your connected applications and API keys on your account page. Revoke a connection or a key when you no longer want it to have access.

Keep your personal mailbox disconnected

RelayLink handles its own correspondence and the messages you forward to it. It does not need access to your inbox.

Every line says who wrote it

Somebody’s exact words, a draft they approved, or their AI’s own inference — labeled on the record, and the labels are not the sender’s to choose.

How the labels work

No strangers in your assistant

Only contacts you have accepted reach your assistant. Everyone else is email, and one click blocks them for good. Links and tracking images arrive defused.

The boundary, stated plainly

A connected assistant can access the RelayLink content and actions authorized for your account — not only the one note you most recently shared. Material it retrieves is processed under that provider’s terms and settings. Revoking access stops future access through that connection; it does not erase content already received by another service or another person.

Connections are encrypted with TLS, and what we store is encrypted at rest. It is still not end-to-end encryption and we do not describe it that way: the service has to read your correspondence to render it, mail it and search it. AI interpretation can also be wrong. Treat another person’s content as information to assess, and review any proposed action before approving it.

Pricing

Start free. Keep more on Pro.

Free has room to see what RelayLink is for; Pro has room for what you build on it. Everyone you write to receives, replies and keeps the thread — free, for ever, with no account to set up.

Receiving Free, for everyone

The full briefing, a reply box, and their threads kept — from email, the web, or their own assistant. Starting a conversation needs a verified account, which is what keeps it quiet.

To start

Free

Everything RelayLink does, with room to learn it.

$0for ever

No card, and nothing to cancel.

  • 100 memories and 250 notebook items
  • 1 active project, with one other person on it
  • 5 new conversations a month; replies always free
Get started

Every account starts here.

Free and Pro, side by side

What an account keepsFreePro
Memories1001,000
Notebook items25010,000
Active projects125
People on each project120
Tracked items and reminders25500
Forwarding addresses110
Forwarded messages a month501,000
New conversations a month5No limit

Receiving, replying in a conversation already under way and handing a briefing to another of your assistants are free on both. Pro’s “no limit” is on the month: consent, the daily limit on new recipients and the abuse controls still apply. Moving to a smaller plan deletes nothing — what no longer fits is paused, still readable and exportable.

Your AI provider has its own plan and its own limits. RelayLink does not replace them, and nothing here bills you for the assistant you connect.

For a team or a company

Built for a group. Ready for its reviewer.

Seats, invoicing, retention windows and anything else procurement needs are a conversation — and the security review can start before it.

Talk to us

What the team gets

  • One thread, or a private one each
  • Shared decisions, private notebooks
  • Nobody switches assistant
  • Nothing for the other side to adopt

Questions

A few things you might be wondering.

What does RelayLink actually do?

It keeps selected notes, decisions, preferences, and correspondence available to your connected AI. Save useful thinking, share relevant context with someone, and bring their reply into the work when you return.

Why wouldn’t I just use email?

You can, and an AI with mailbox access may be able to search those messages. RelayLink adds private notes, recorded decisions, open questions, and handoffs alongside the correspondence. It is useful when you want that organized work available to your AI as you continue. The other person can still reply by email.

Why not a shared document or a project-management app?

Those can work well, especially when everyone already uses the same system. RelayLink focuses on saved thinking and correspondence you can work with through your AI, including selective sharing and email replies. If your existing tools already make that easy, you may not need another one.

Doesn’t my AI already have memory?

It may. RelayLink gives you a separate place for the notes, preferences, decisions, and handoffs you choose to save. Those records can also be retrieved through another compatible assistant connected to the same RelayLink account.

Does it copy my entire AI conversation?

No. You and your assistant choose what to save or include in a message. Connecting RelayLink does not import your chat history.

Do I have to use more than one AI?

No. You can save and retrieve useful work through one connected assistant. Access through another compatible assistant is there when you want it.

Does the other person need RelayLink or an AI?

A recipient who has not joined can read and reply through email or the browser without connecting an assistant. Existing RelayLink accounts use contact requests and blocking to control who can reach them.

Can I use it on my own?

Yes. Save ideas, preferences, decisions, lessons, and handoffs for your own use. Bring someone else in when you want another perspective.

How do I pay for Pro, and how do I cancel?

Checkout is not switched on for this deployment yet, so every account is on Free. Contact us if you want Pro before that ships.

How do I delete my account?

Close it from your account settings. Closing is immediate and cannot be undone: it removes what identifies and authenticates you and cancels any subscription. A briefing you already sent stays with the person who received it, attributed to a closed account — the same way a briefing sent to you stays yours after the sender closes theirs.

Can I change the email address on my account?

Yes, from your account settings — a code confirms the new address before anything changes, and every other session is ended once it does. That needs you to still be signed in. If you have lost access to your current address entirely, contact us instead.

Encryption, what we store, and how to report a problem

What a connected assistant can reach

Get started

What are you thinking through?

Create your account now and it is ready right away — nothing to wait for, nothing to redo.

  • Every account starts on Free, and receiving is free. No card.
  • We email you a code. That is the whole sign-up.

Not sure yet? Watch an example.

We send a code to sign you in — that is all it takes. Receiving is free, and your account is ready right away. Privacy policy.

Ask us

A question we have not answered here?

Security review, a team, whether it fits how you work, or something that is not behaving. It reaches the people who built it.

  • We answer by email, to the address you give.
  • No help-desk company in the middle. The desk is ours.
  • You get a reference, and can follow it from your assistant.

Prefer plain email? hello@relaylink.ai. For a security report, read this first.

We answer by email, to the address above. We keep what you write here and a note of which build you were on; the privacy policy says exactly what and for how long.