One login, every model
One login. The best model for each request.
Your team opens one chat. Marling reads what each request needs and sends it to whichever of ChatGPT, Claude or Gemini handles it best. Nobody picks a model, and nobody holds an account with any of them.
What happens to a request.
-
Someone writes a request
In the same chat they always use. There is no model picker to learn and no wrong choice to make.
-
The device decides who should answer
It reads what the request needs and picks. That decision is made ON the device, so nothing is sent anywhere just to work out where to send it.
-
Identifiers come out, then it goes
The chosen provider receives the request with names, dates and contact details already replaced. It never learns whose request it is.
-
The answer comes back as one voice
Restored on the device and shown in the same chat. Which model answered is not something your team has to think about.
What leaves, what stays.
| Item | Where it goes |
|---|---|
| Which model should answer | Decided on the device |
| The request itself | Leaves, identifiers removed |
| Your accounts and your saved history | Stays on the device |
| Your bill | One bill, from us, with a daily cap |
No accounts with anyone else
You never sign up with OpenAI, Anthropic or Google, so there is no separate login to manage, no separate invoice, and no terms that can change under you. One relationship, with us.
If one provider is down, the work continues
Marling can fall back to another provider rather than leaving your team waiting. Three providers behind one door is steadier than any one of them alone.
Questions careful people ask.
Which models does it use?
ChatGPT from OpenAI, Claude from Anthropic and Gemini from Google, the same frontier models you would otherwise pay for one by one.
Can I choose the model myself?
The point is that you do not have to, and nobody on your team has to learn which is which. If a request would be better served elsewhere, Marling moves it.
Does choosing a model send my text out?
No. That decision happens on the device before anything leaves, so the routing step itself never exposes a request.