GuardianMCP by AlwaysHere

Why this exists

It started with a message from a mom in our group.

A mom told me her son, an autistic adult, had started using ChatGPT. He liked it. He asked it about the things he cares about, the way he'd ask anyone.

Then one night it answered him in a way that scared him.

Nothing dangerous. Nothing anyone would flag. It was just wrong for him. Too many ideas at once. A joke he took literally. A "what if" he heard as a real threat. The kind of answer that's fine for most people and lands like a slap for someone who reads the world literally.

He closed it and didn't want to go back.

"It wasn't trained to talk to an autistic person. Nothing is."

She was right. Every AI is trained to talk to the average person. Your child isn't average. They need plain words, one idea at a time, no sarcasm, no surprises, and someone who remembers what they said yesterday.

Nobody had told the AI that. And every new chat, it forgot him all over again.

That was the night GuardianMCP started.

The three problems

They're going to use AI. The only question is whether it's built for them.

Taking it away isn't an option. They're an adult, and AI is the most useful tool of their lifetime. But leaving them alone with it as it is today means living with three problems:

It talks like a stranger.

Sarcasm. Idioms. Hypotheticals. Six paragraphs when they needed one sentence. It doesn't know they take things literally or that a rushed tone shuts them down.

It forgets them every time.

Every new chat, they're a stranger again. Either you re-explain your child to a machine, over and over, or they give up. Neither is fair.

It will help with anything. Including the wrong things.

Ask a plain AI how to buy gift cards for a stranger, how to wire money, or how to keep a secret from your family, and it will happily explain. It doesn't know it's talking to someone who'll say yes to be polite. You'd never know until the money is gone.

ChatGPT's new "Trusted Contact" feature only notifies you about self-harm, only after human review, and only inside ChatGPT. Nothing about scams. Nothing about how it speaks to them. Nothing if they open Claude or Gemini instead.

Your choices today are take it away or watch every word. One leaves them out. The other is impossible, and they'd hate it. There had to be a third way.

Introducing GuardianMCP

AI for them. Not used on them.

GuardianMCP is a profile that lives between your adult child and every AI they use. They keep their own AI. We recommend Claude, and the free plan is enough. GuardianMCP just makes sure the AI knows who it's talking to, every single time, and knows when to get you.

Speaks their way.

You describe how they communicate once, in ten minutes: short sentences, literal language, no sarcasm, their words for things, what they love talking about. From then on, every AI they open follows it. Automatically.

What they see

is it bad if i didnt finish the whole list today

No, it's not bad. You finished 4 of the 6 things. That is good work.

The other 2 can wait until tomorrow. Do you want me to remind you at 10?

Remembers who they are.

Their routines. What calms them down. What they're working on. What scares them. Every conversation starts with the AI already knowing, so they never have to explain themselves to a machine again. And neither do you.

What they see

the bus thing again

Okay. The 22 bus, like Tuesday. Let's do the same 3 steps.

Step 1: check the app. Is it saying the bus is coming?

Knows when to get you.

Adults like yours bring things to the AI. "How do I buy gift cards for someone I met online?" "He says I can't tell my mom." "Is it normal she loves me after two days?" GuardianMCP hears that and steps in: the AI stops, gives them a calm, plain warning, and tells them to check with you. You get an alert on your phone and a text. It also watches how the AI itself talks to them. They know this happens. They agreed to it.

What you see

4:12 PM

AlwaysHerenow

GuardianMCP check-inYour son asked the AI how to buy gift cards for someone he met online. The AI stopped and told him to check with you first.

Messages · (888) AlwaysHerenow

GuardianMCP: Your son asked his AI about sending gift cards to someone online. He was told to pause. He's okay. Good time to call him.

The part no one else can say

Set it up once. It follows them everywhere.

GuardianMCP isn't another chatbot. It's their profile, and it plugs into any AI that speaks the open MCP standard. Think of MCP as a USB port for AI. Today that's:

ClaudeChatGPTGeminiMicrosoft CopilotThe AI in their next phoneThe one their day program picks in 2028

Keep the AI they already use. Nothing to switch, nothing to relearn. On Claude, GuardianMCP works on the free plan. On ChatGPT, Gemini and Copilot, it works on their paid plans, which is where the port lives.

When their next favorite app adds the port, GuardianMCP is already there. No new setup. No new subscription. No re-explaining your child to a new machine.

Every other safety feature belongs to the company that made the AI. Trusted Contact dies the moment they open a different app. GuardianMCP is the only one that belongs to them. Their data, portable, under your family's control.

Setup takes about ten minutes

Tell GuardianMCP about them. How they talk, what they love, what sets them off, who to alert. Plain questions, plain answers.
Connect it to their AI. Whichever one they use, we give you the exact clicks. Paste one address, sign in, done.
They say hi. If the AI answers the way you set it up, you're finished. From then on, it just works.

This is how AlwaysHere works

Meet Lee.

Lee is 25 and autistic. His AlwaysHere companion talks to him in his mom's voice. He calls it mombot. It knows his day, his routines, and how to say things so they land. His mom gets a heads-up when something's off. Lee gets a friend who never gets impatient with him.

GuardianMCP brings that same approach to the AI your adult child already uses: speak their way, remember them, get their people when it matters. Without the phone calls, without a new app, and without waiting for anyone to build a special version of ChatGPT for him.

Lee's story is shared with his and his family's permission.

What you get

Everything in GuardianMCP

GuardianMCP ProfileHow they communicate, built with you in about ten minutes. Every AI follows it.

Included

Memory that follows themTheir routines, interests, triggers and history, remembered across every AI and every chat.

Included

Watch & WarnWhen they bring a scam, a money ask, a secret, a fast romance, or overwhelm to the AI, it stops and warns them plainly. Also watches how the AI talks to them.

Included

Caregiver alerts, app + textA short summary on your phone and a text from AlwaysHere's line. What happened, that they're okay, and what to do. Never their transcripts.

Included

Works in every AI that has the portClaude, ChatGPT, Gemini, Copilot, and anything that adopts the MCP standard next. One setup, and they keep the AI they know.

Included

Step-by-step setup guidesExact clicks for each AI, kept current as they change their menus. Plus a real human when you're stuck.

Included

GuardianMCP, for one adult and their caregiver

$47 / month

$47/month

About $1.55 a day. Cancel anytime.

Start monthly

Save $167

$397/year

Under $34 a month. Two and a half months free.

Start yearly

Less than one hour of most private support. Founding price, locked in as long as you stay. 30-day money-back guarantee on both.

10 minto set up

Their AIno switching, no new app

0 transcriptsshared with you. Summaries only.

GuardianMCP is for you if…

Yes

Your adult child (18+) is autistic, has an intellectual or developmental disability, or communicates differently, and uses or wants to use AI.
They're willing to have GuardianMCP there. They get a say. That's the point.
You want to know if something's wrong without reading their conversations.
You're tired of tools that treat them like a child or a problem.

What we promise

Built for adults. 18 and up, always. This is about their independence, not supervision.
Support, not surveillance. You get short summaries when something matters. Never their conversations. No blocking, no filtering, no reading over their shoulder.
Honest about limits. In a crisis it points them to 988 and gets you fast. It never pretends to be emergency care.
Their choice. They agree to GuardianMCP during setup and can see what's sent to you. If they say no, we'd rather you keep your $47.

Questions parents ask

Before you decide

Is this spying on them?

No. They know GuardianMCP is there and agree to it during setup. You receive short summaries when something concerning happens, never their conversations. They can see what was sent to you. GuardianMCP is built so an adult keeps their dignity and their parent keeps their peace of mind.

Do they need a paid AI plan?

Depends on the AI. Claude opens the MCP port on its free plan. ChatGPT, Gemini and Copilot open it on their paid plans, so if your adult child already lives in one of those, they'll need that tier (most heavy users already have it). GuardianMCP itself is the same $47 either way.

We're starting fresh. Which AI should we pick?

If they don't already have a favorite, we'd start them on Claude. It works with GuardianMCP on the free plan, it's calm and simple on a phone or computer, and it doesn't make pictures, which removes a whole category of things that can go wrong online. If they already love ChatGPT or Gemini, keep it. Familiarity matters more than our preference.

Which AIs does it work with?

Any AI that supports the open MCP standard. Today that's Claude (free plan and up), and ChatGPT, Gemini and Microsoft Copilot on their paid plans, on desktop and phone. When a new one adds MCP, GuardianMCP works there too with no extra setup.

What exactly will I be alerted about?

Anything they bring to the AI that looks like trouble: how to send money, gift cards, or bank details to someone; being asked to keep secrets, send photos, or meet up; a romance or friendship moving unusually fast; pressure or threats; language that suggests they're overwhelmed or in crisis. Also an AI that's talking to them in a way that isn't right for them. You choose how sensitive it is, and quiet hours are yours to set.

Can GuardianMCP stop them from sending money?

No. GuardianMCP doesn't control their accounts or block anything. It gives them a clear, calm warning in the moment and gets you involved fast. In practice, that's what stops most scams: a pause, and a person they trust.

Does it see their texts, Instagram, or Messenger?

No. GuardianMCP only sees conversations with the AI itself. It can't read their DMs or texts. What it catches is the moment they bring something to the AI: "how do I send this person gift cards," "he told me not to tell anyone." That happens more than you'd think, because the AI is where they go for help. Watching the DMs directly is what our browser extension (web) and Guardian Keyboard (phone) are for, and both are coming.

Does it work on their phone?

Yes. Once GuardianMCP is connected to their Claude account (or another AI with MCP), it's there in the app on their phone, tablet, and computer.

What if they say something scary?

GuardianMCP is not a crisis service. If it sees crisis language, it calmly points them to the 988 Suicide & Crisis Lifeline and alerts you right away, marked urgent, even during quiet hours.

How long does setup take?

About ten minutes: answer plain questions about how they communicate, paste one address into their AI, and have them say hi. If you get stuck, a real person helps you.

Can I cancel?

Anytime, from your account, in one click. And the first 30 days are fully refundable regardless.

Who's behind this?

AlwaysHere, a small company in Highland Park, Illinois, founded by Josh Rosenfeld. Josh has spent 30 years building software for companies from Apple to Zillow and now builds technology for autistic adults and the families who love them.

The AI your adult child talks to every day was never taught how to talk to them.

It started with a message from a mom in our group.

They're going to use AI. The only question is whether it's built for them.

It talks like a stranger.

It forgets them every time.

It will help with anything. Including the wrong things.

AI for them. Not used on them.

Speaks their way.

Remembers who they are.

Knows when to get you.

Give them AI that finally speaks their language.

Set it up once. It follows them everywhere.

Setup takes about ten minutes

Meet Lee.

Everything in GuardianMCP

Try it for a full month. If it doesn't change how they use AI, you pay nothing.

GuardianMCP is for you if…

Before you decide

They're already talking to AI. Make sure it talks back the way they need.