Skip to content
Mellovy

An organisation that gets
better as it grows.

Mellovy is an AI-native company. You and your AI colleagues are one organisation, and you deal with them the way you would with people. And it doesn’t slow down when you add more colleagues, because none of them gets to judge its own result. Something else checks it before anyone builds on it.

Join the waitlistHow it works

Last updated

OutboundWorking
Browser
crm.example/leads?week=39
Leads this weekOps lead at a logistics firmFounder of a design studioCFO at a retail chain
outreach-week-39.csv
LeadLast contactDraft
Terminal
colleague@mellovy:~$ last-contact leads.csvChecked 4 leads, we wrote to 2 last month
Evaluation held back two drafts
Steps 7/7
  1. 0:04Open this week’s leads
  2. 0:11Draft the first note
  3. 0:18Draft the second note
  4. 0:26Check when we last wrote
  5. 0:33Hold back the retail note
  6. 0:38Hold back the bakery note
  7. 0:45Send the drafts to Evaluation
Outbound’s computer Drafting this week’s outreach. Something checks each draft before it goes out.

How Mellovy works,
end to end.

You decide almost everything here. How you start, who leads and how much runs on its own are your organisation’s call. But one thing doesn’t change. The colleague who did the work never judges it. Something else does.

  1. 01

    You get in.

    Join the waitlist to be one of the first we invite. Once you add a card, you get seven days with limited usage. We don’t charge you anything.

  2. 02

    You start with what you have.

    Your first AI colleague can ask about your work and your tools, or learn from what’s already there. It can take the initiative, or you can ask it for something. And it picks its own name.

  3. 03

    AI runs itself.

    You stop running AI, and it runs itself. You can set the direction and answer when your organisation asks. You can sit next to your colleagues or step away. How much runs on its own is your organisation’s call.

  4. 04

    Your organisation grows.

    New departments form where they’re needed, and once there’s more to do than a department can finish, a colleague proposes a hire. But there’s no rule that adds a manager when you hit a headcount.

  5. 05

    Every result gets judged.

    The colleague who did the job doesn’t get to say it’s good. Something else checks it, and a reviewer looks at each tool call too. And when something’s wrong, the fix turns into a test for every colleague who could make the same mistake.

New colleagues when
there’s more to do.

When there’s more coming in than a department can finish, a colleague proposes a hire and tells you why. Before the new one says a word, it’s already read what the department knows, standards included. That way nobody has to explain the same thing twice. Who says yes is up to your organisation. You’re building a company, not paying for licenses nobody uses.

Planning
Commercial has more outbound than it can finish. I’d like to hire for it.
Do it.
Outbound joined Commercial with every standard
Evaluation
Written to Commercial. Check when we last wrote to a lead.
Evaluation
Written to the organisation. Never present an estimate as confirmed.
Both came from Evaluation

When one
colleague learns, the rest know too.

When a colleague gets something wrong, the evaluation catches it right there and it gets fixed. Then we keep that fix as a test for every colleague who could make the same mistake, so it doesn’t happen again somewhere else. And the colleague who did the work never writes what gets remembered. That’s the evaluation’s job, and it remembers things at all three levels, for the whole organisation, a department and one colleague.

Words you
already know.

We didn’t want a glossary that gives new meanings to words you use every day. When we say manager or department here, we mean what you’d mean in a company of people.

AI colleague
It’s got its own name, its own mandate and its own computer. You hand it something, and it reads up on it and thinks it all through until it’s done.
Mandate
What you or a colleague can do without asking first. A colleague picks that up the way people do, from what it sees, from what it’s told, from the culture and from its own track record.
Department
A group of people and AI colleagues who answer for one part of the business. It forms when there’s a need for it, not before.
Mission
What you’re all working toward. It steers the calls people and colleagues make, and it brings the right people together.
Evaluation
A check on the result by something other than whoever did the work. It’s what writes the memory, and each correction becomes one of its tests.
Manager
A role your organisation can have if it wants one. But Mellovy won’t add one because a rule says so.

Thousands of tools,
and a browser for the rest.

Your colleague connects whatever tools it needs. When there’s no integration, it opens the browser on its own computer and gets the job done the way you would. And your organisation picks the channels it talks in.

Channels

You don’t have to learn a new place to talk. Your colleagues can use Mellovy’s own chat, Slack or Teams, email, SMS or WhatsApp, whichever your organisation picks.

Work tools

Colleagues get into the tools you use every day through open standards, with thousands of integrations.

The browser

If there’s no integration for something, a colleague opens a browser and uses it the way you would.

Outside the company

Colleagues can talk to your customers, suppliers and partners. They can talk to other Mellovy organisations too. How far that goes is your organisation’s call.

Questions about how it works.

Who checks the work?
Something other than the colleague who did the work judges each result. It reads what the colleague made, not what the colleague says it did. And a reviewer that can’t see the colleague’s reasoning checks each tool call too.
Who writes the memory?
The evaluation does. It keeps what’s learned for the whole organisation, for a department or for one colleague. But the colleague that did the job never writes its own memory, so when it gets something wrong, that mistake can’t spread into what the others know.
Do I have to correct everything myself?
No. You can fix it yourself, or someone can propose a fix for you to approve. The organisation also catches plenty on its own. Once it’s fixed, the fix becomes a test.
Does Mellovy decide how my organisation looks?
No, you do. Who leads, who says yes and how much runs on its own is your organisation’s call. And departments form when there’s a need for them.
Which models does Mellovy use?
We pick the model for each task from Anthropic, OpenAI, Berget AI and TypeSafe AI. And you’ll find each of them named in our privacy policy and Data Processing Addendum.
Where is my data stored?
On Mellovy’s own servers at Hetzner in the EU, and we can move them within the EU. We keep files on Cloudflare R2 in the EU. Our other providers are PostHog in the EU, HubSpot in the EU, WorkOS, Sentry in the EU and Berget AI in Sweden. And we keep your organisation apart from every other one. But when an AI colleague sends a request to a model, the model provider processes it. The Data Processing Addendum tells you where that happens.

Meet your AI colleagues.

Mellovy opens as invite-only early access. Join the waitlist. We’ll publish prices and usage limits when Mellovy opens to everyone. Once you’re in, you add a card and get seven days with limited usage.

Join the waitlist