Start here

Two directions.
Most people mash them together.

There are two wires here. They do different jobs, and they travel opposite ways.

One brings your AI to DreamerOS. The other lets DreamerOS reach into the accounts you already own.

Tell them apart and the rest of this page is easy. So here they are, plainly, including the boundaries we hold on purpose.

The two wires, side by side

Which way	What moves	What it costs
Coming in	Your AI tool connects to DreamerOS and borrows a memory, a check, and a receipt.	Free.
Going out	DreamerOS connects to your own accounts. It reads from many of them and writes to eight.	Depends on your tier. Your account page confirms it.

These are separate. Plugging Claude into DreamerOS does not give DreamerOS your GitHub. You grant that on its own, one account at a time, and you can take it back.

Start with the wire coming in

Direction one, coming in

Your AI connects to us.
That wire is free.

Claude. ChatGPT. Cursor. VS Code. Codex. Windsurf. They do the thinking. That part is theirs and it stays theirs.

DreamerOS hands them three things they do not have on their own. A memory that lives outside them. A check on the answer. And a receipt for what happened.

Here is what that feels like. Decide something in Cursor on Tuesday. Ask ChatGPT on Thursday. It already knows.

This wire is free, and we think it should stay free. You should not have to pay to stop repeating yourself.

A memory that lives outside the tool

Every AI tool you open starts over. It has never met the last one. So you re-explain the project, the people, the way you like things written. Five times a day, in five tabs.

The memory sits with DreamerOS instead of inside any one tool. Each tool reads the same memory. So the tools stop being strangers to each other.

The useful part is what this does to the model question. The model is the swappable bit. The knowing is not. Bring whichever one you like this month and it arrives already knowing you.

Why we keep this side free

Plain answer. A memory you cannot try is a memory you cannot trust. Charging at the door would mean asking you to believe us first and check afterwards, which is backwards.

So connect the AI you already pay someone else for. Use it. If the going-out wire is worth money to you later, that is a separate decision you make with evidence in hand.

Set up your AI tool

How do we prove that, out loud?

Checked, not promised

Do not take our word for it.
Read it yourself.

The wire coming in is not a sales pitch. It is a public document, and any AI system can read it in one request.

That document names every real way in. Four ways work today. An older, fifth way is listed too, marked broken, because we do not hide the things that do not work.

The same document used to undersell what DreamerOS can do. That is fixed now. Honesty runs both ways. We do not round ourselves up, and we do not round ourselves down.

What the document actually says

Checked live against mcp.dreameros.app. Four connection doors are listed as working: the standard door most AI tools already use to connect, a door for OpenAI-style clients, a matching door for Anthropic-style clients, and a plain web-request door with its own published reference.

A fifth entry sits in the same list: an older way in. The document says outright that a client pointed at it waits forever and never connects. It stays listed anyway, because automated scanners still probe for it, and removing it would hide the failure instead of naming it.

The document also states how current DreamerOS says it is. Until recently, it named an older, weaker level than the software behind it already supported. Nobody was claiming that to look good. DreamerOS was understating its own reach. Both are stated correctly now.

Anyone can check this without asking us first. A competitor, an auditor, or the AI tool you already use can fetch the same document and see the same list.

Read the document yourself

Now the wire going out

Direction two, going out

We reach into your tools.
We read from many. We write to eight.

Reading is wide. Writing is narrow. That gap is deliberate.

Eight actions can change something in an account of yours. Eight. Here they all are, with nothing rounded up and nothing left out.

A ninth one drafts email and stops. It never sends. More on that below, because it is a promise rather than a limitation.

The eight writes, all of them

GitHub. Create an issue.
Notion. Create a page.
Linear. Create an issue.
Linear. Move an issue to a new state.
Linear. Leave a comment.
X. Post.
Slack. Send a message.
Resend. Send an email.

That is the list. If you can think of a ninth thing you would like it to do, it does not do that thing yet. We would rather you read a short honest list than a long hopeful one.

Gmail drafts. It never presses send.

Gmail is the one that stops short on purpose. It writes into your drafts folder and it stops there. There is no send in the code at all. Not switched off. Not there.

Read that as a promise rather than a gap. The last move stays yours. An AI that can send email on its own, in your name, while you sleep, is a bad night waiting to happen.

So you get the writing done for you and you keep the deciding. You open the draft, you read it, you press send. Or you do not.

What it reads

Reading changes nothing on your side. It is how the answer gets to be about your actual work instead of about work in general.

Google Calendar. Reads your schedule. It does not create events. See the boundaries beat below.
Google Drive. Your files.
Stripe. What is being paid.
Sentry. What is breaking.
Railway. What is running.
Vercel. What shipped.
Cloudflare. What is in front of it.
Airtable. Your records.
Porkbun. Your domains.
Supabase. Your data.
Firecrawl. Pages from the open web.

You grant each one on its own. Nothing here turns on because something else did.

See the connect surface

What the combination gives you

Both at once

One wire is useful.
Both wires is the point.

Each wire on its own is a convenience. Together they are the thing you cannot get anywhere else.

Here is the plainest version. It reads the error in Sentry. Then it files the issue in GitHub. You did not carry anything between the two.

Five things that only work when both wires are live

It reads, then it acts. The Sentry error becomes the GitHub issue. One motion. You were not the courier.
A decision travels. Decide it in Cursor. It shows up in the Notion page later. Both ends read the same memory, so nothing had to be pasted.
The update knows the story so far. The Slack message goes out already knowing what the team was told last time. It does not repeat itself and it does not contradict itself.
The model stops mattering so much. Bring whichever one you like. It arrives already knowing you. The model is swappable. The knowing is not.
A Zap can ask first. An email lands, a Zap asks DreamerOS what you decided about that client, and the reply is drafted your way. That is the Zapier wire coming in, and it works today.

The boundaries we hold on purpose

The boundaries

What it deliberately does not do.
Said out loud, so the rest can be trusted.

This is the part most pages leave off. We think it is the part that makes the other parts worth reading.

Nothing below is coming next week. When one of these changes, this beat changes with it, and not a day earlier.

The six honest limits

It does not send email. Gmail is drafts only. That one is a choice we made and intend to keep.
It does not write to your calendar. It reads your schedule. It cannot put anything in it.
It cannot message you first. There is no push and no helpful nudge. You start every conversation. It never starts one.
Below Elite there are no teams and no shared workspaces. One person, one memory. Colleagues cannot read yours and you cannot read theirs. Elite adds team management, roles and shared outputs, and even there each person keeps their own memory and their own record.
Much of the connect catalog says coming online. Coming online is not connected. A name on that page is a plan, not a wire, until it says otherwise.
Zapier works one way today. Your Zaps can call DreamerOS. DreamerOS cannot yet start your own Zaps from your own Zapier account. Until it can, that card says coming online.

If any line on this page ever drifts from what the product does, this beat moves first and the selling moves second. That order is the whole point.

Try the free wire

Start

Bring the AI you already use.
See what it feels like to be known.

The coming-in wire is free, so start there. Connect the tool you already have open.

Ask it something on Tuesday. Ask a different tool on Thursday. If Thursday already knows about Tuesday, you have felt the whole idea.

Then, and only then, decide whether the going-out wire is worth anything to you.

Where to go next

Set up your AI tool walks the paste, one step at a time, for the client you actually use.

Connect lists the providers and what each setup path looks like. Your account page confirms what your tier can use before you grant anything.

Intent fidelity is the standard behind the check on the answer, if you want the deeper read.

The live receipt ledger

Connect your AI, free

Plain words for the words on this page

Page. A single screen on this site, like the one you are reading. It is not a unit you buy and we never charge for one.

A client-side path in the web app, served over HTTPS; not a billing unit in any plan.

Token. The small pieces a computer chops your words into, about three quarters of a word each. Your plan sets how long one answer can be, counted in tokens. It does not set how many questions you may ask.

Max tokens per response, one value per plan. Each plan card lists its own.

Engine. One AI brand you can talk to. ChatGPT, Claude, Gemini, Grok and Perplexity are the five. Your plan page lists which ones you can use.

Five role names per plan (Build, Review, Synthesis, Signal, Research). Some plans count the engines you connect live, and the limit is checked on every request.

Connector. A link between DreamerOS and a tool you already use, such as Gmail, Notion, Slack, GitHub or Stripe. It lets DreamerOS read from that tool, and in some cases act in it, but only inside rules you set.

38 provider integrations on the gateway, each reached over HTTPS with a scoped, per-provider key.

Receipt. A short record attached to your answers. It names which engine answered, what it cost, which checks passed, and whether sources were checked. You never have to guess what just happened.

One record per answer, written server-side on the gateway: engine, model, plan, cost, and the pass or fail result of the pre-check and post-check.

Memory. The facts about you that DreamerOS keeps and reuses next time, so you stop repeating yourself. Your plan sets how many facts it holds.

Active memory rows per plan, with a smaller fallback for a session with no plan set yet. Each plan card lists its own.

MCP key. A key you paste into another app, such as Claude Desktop or Cursor, so DreamerOS memory and checks work there too. You do not have to change where you already work.

A signed key. Individual tools inside it may carry a plan floor.

Intent fidelity. Checking that the answer answers the question you actually asked, not a nearby one that was easier.

A rule-based check that runs on the gateway, scored on three axes, with no separate model call.

Direction set. The standing written instructions DreamerOS gives every engine before it sees your question, so all five behave the same way.

One canonical system prompt, loaded by the gateway and sent to every engine.

Check. A test run on your question before the engine sees it, and on the answer before you see it.

A pre-check before the engine sees your question, and a post-check before you see the answer; both results ride the receipt.

Two directions. Most people mash them together