Onsay Start free

Software that fixes itself

Your users say what is wrong. The software changes. You approve it, or you tell it to stop asking.

Onsay reads what your users tell you, works out which of it is worth doing, and writes the change.

Free to start. No card.

A user said

“every time i log in on my phone it signs me straight back out. happens constantly, i have basically stopped using it on mobile”

Onsay decided

People are being signed out on mobile

Worth doing 95% Your customer 85% A real request 95%

Onsay wrote this

- maxAge: THIRTY_DAYS,
+ sameSite: 'lax',
+ secure: true,

You approve it before any of it ships

Someone tells you the export button is impossible to find. You agree with them. You write it down somewhere, or you do not, and either way it is still impossible to find four months later when the next person says the same thing.

The problem was never knowing what to fix. You knew. The problem is the distance between somebody saying it and it being true.

Onsay closes that distance. What people tell you comes back as a decision you can make in ten seconds, and the ones you approve get built.

Start to finish

One person said one thing

They said

“every time i log in on my phone it signs me straight back out. happens constantly, i have basically stopped using it on mobile”

It came back decided

People are being signed out on mobile

Worth doing 95%

A signed in person being signed out is a broken core path. Nothing else on the list matters while this is live.

Your customer 85%

An existing paying customer, on the platform most of them use.

It shipped

Keep the session across the sign in redirect on mobile Safari

Three people described the same thing in different words, so it became one change. Sets the session cookie flags correctly and adds a test that fails if this breaks again.

Approved on a Tuesday. Live the same afternoon.

And this is what landed in the code

server/lib/session.js onsay/auth-mobile-session
1 . export const startSession = (res, token) => {
2 . res.cookie(SESSION, token, {
3 . httpOnly: true,
- maxAge: THIRTY_DAYS,
4 + maxAge: THIRTY_DAYS,
5 + // Safari drops the cookie on the redirect back from sign in
6 + sameSite: 'lax',
7 + secure: true,
8 . })
9 . }
Pull request opened #215 3 files changed Not merged. That part is yours.

What comes back

Not a summary. Not a sentiment score. A decision, with the reasoning attached.

It says what the person wants, and whether they are serious, because people joke, test and vent. It says why they want it, if they said. It judges whether your product actually needs it, measured against your own description of what your product is for rather than against how loudly they asked. And it judges whether the person asking is the customer you are building for, because sometimes they are not, and building for the wrong person is worse than building nothing at all.

Then it says what it would change, and roughly where.

You read three numbers and a paragraph. You say yes or no. That is the entire review.

Two gates, both yours

The first gate is the decision. Nothing is touched until you accept it. Reject it and it is gone.

The second gate is the change itself. Onsay reads everything people said together, finds the gap underneath, and proposes a change that closes it, often one idea that answers several requests at once rather than five disconnected fixes. Approve it and it is built on your schedule. Weekly by default. Or immediately, or monthly, whichever suits how you ship.

Nothing ever goes live on its own. The work arrives finished and waits for you.

Why the judgement matters more than the code

Every tool that writes code from a description has the same problem. The description is the hard part. Somebody still has to decide what is worth doing.

That decision is the thing Onsay is actually for. When four people ask for the same thing in four different ways, that is one piece of work and not four items on a list. When somebody asks for an enterprise integration and you sell to freelancers, the useful answer is no, with the reasoning written down so you can spot it if it keeps happening.

You see the reasoning before a line is written. That is the whole reason it exists in that order.

Getting what people say into it

  • A small box on your site. One line to add it, and it cannot collide with your styling.
  • Paste it in yourself. Something from a call, an email, a message someone sent you at eleven at night.
  • Send it from anything that can make an HTTP request. Point your support inbox at it.

You do not need the box to start. Paste in the three oldest things anyone has told you and see what comes back.

What we keep

What we need to make the decision and do the work. We do not keep a copy of your source. The box takes an email address only if the person types one, and only so you can reply to them.

One that already happened

Something a person actually said, and what it turned into.

every time i log in on my phone it signs me straight back out. happens constantly, i have basically stopped using it on mobile
what came back sign in half a day

People are being signed out on mobile

Worth doing 95%

A signed in person being signed out is a broken core path. Nothing else on the list matters while this is live.

Your customer 85%

An existing paying customer, on the platform most of them use.

A real request 95%

Describes a pattern they can reproduce, and a change in their own behaviour.

What they want
To stay signed in on their phone. This is a report of something broken, not a request for anything new.
Why it matters to them
They have stopped using it on mobile. That is a customer leaving slowly.

What would change

The session is dropped on the redirect back from sign in, which points at the cookie flags rather than the login itself. Fix the flags and add a check so it stays fixed.

shipped 3 files

Keep the session across the sign in redirect on mobile Safari

Three people described the same thing in different words, so it became one change. Sets the session cookie flags correctly and adds a test that fails if this breaks again.

Approved on a Tuesday. Live the same afternoon. You approved it first. Nothing ships without you.

Questions

Does it change anything on its own?
No. Work arrives finished and waits for you to approve it. Nothing goes live without you.
AI code is garbage though.
You see the reasoning before a line is written. If the reasoning is wrong you reject it and nothing gets generated at all.
What people tell me is in Slack and email, not a widget.
Then paste it in, or send it from whatever you already use. The box is one of three doors and it is the one you need least.
What does it cost?
Nothing to start. Feedback, docs and your roadmap are free on every plan. Paid plans start at $9 a month and cover the steps Onsay ships.

Your users say what is wrong. The software changes. You approve it, or you tell it to stop asking.

Free to start. No card.

Try it on something a user told you

Anything real. Nothing is saved.

Start free