Platform staff only.
Overview
Platform-wide customers, users and content.
Platform
Spend and margin
Daily trend
By provider
Usage
Consumption by organization, seat, feature and model, down to the individual call.
Costs & margin
What the platform pays providers, what customers are charged, and the gap between them.
Overage to bill
Generations used past the monthly allowance this month, and what to charge for them on top of the plan.
By provider
By model
Unpriced models
Used in the last 90 days with no rate card entry, so billed at zero.
Organizations
Manage customer workspaces, seats and entitlements.
| Customer | Status | Seats | Quota | Content | Actions |
|---|
Users
Search accounts across customers; force logout or reset.
| User | Customer | Role | Status | Actions |
|---|
Rate card
Provider cost and customer price per model. Versions are effective-dated and never edited in place.
Customer pricing
What each product action is charged at, independent of provider tokens. Set a flat price per output, or a markup on cost. A flat price always wins.
Provider cost
Provider cost and customer price per model. Versions are effective-dated and never edited in place.
AI assistant
What the AI assistant is allowed to do, what it consumed, and a record of everything it did. Permissions are enforced when a tool runs, not by asking the model to behave.
Assistant permissions
Everything is denied unless granted here. A capability that is off cannot be reached however the assistant is asked for it.
Customers configured differently
Workspaces whose assistant permissions do not match the platform default. Everyone not listed here inherits it.
Opening reply
What the assistant says before the customer has typed anything, and the quick-start options it offers. Write it yourself, or let the model write it within the rules you set here.
Where the spend goes
Conversation tokens against what the assistant's tools generated on the customer's behalf.
Daily cost and charges
Assistant pricing
The markup applied to what the assistant costs us, and the per-customer multiplier on top of it. Published as a new version — existing charges keep the price they were billed at.
By customer
By user
By model
By capability
Which tools the assistant reached for. Pairs with the cost above: a spike is only actionable once you know what caused it.
Every tool call, refusal and detected probe. Arguments are truncated and stripped of anything credential-shaped before being stored.
Model providers
Connect a provider once with its API key. Every model it offers — text, image, video, music — becomes assignable to a feature below.
Models by feature
Which provider and model each product feature generates with. Text features take any connected provider; media features need a generator we have built for that vendor, and any provider that cannot serve a feature is listed with the reason rather than hidden. Workspaces that bring their own model keep it for text features.
Customer-owned providers
Customers you allow can connect their own model endpoint and pay their provider directly. Turning this off returns them to the platform provider on their next generation — nothing they configured is deleted.
Audit log
Immutable record of every privileged action.
| Time | Action | Actor | Customer | Detail |
|---|
My account
Your operator profile, password and two-factor authentication.
Profile
Change Password
Changing your password signs out every other device.