AGENTRA

routerPlex

A control point between your team and the AI providers: quotas per person, which model each one may use, and how much was actually consumed.

Govern how much AI your team consumes

The problem

An organisation starts paying for artificial intelligence and two months later has no idea who is consuming what. Provider plans offer no control per person or per device: there is a global limit and a bill at the end of the month. By the time someone overruns, they have already overrun.

How it solves it

routerPlex sits in the middle. All traffic to the provider passes through it, gets identified by team, is checked against that team's quota and its list of permitted models, and only then is let through. Control is applied before the call, not after the invoice.

How it solves it

Every team’s traffic passes through routerPlex, which checks identity, quota and permitted models before letting it reach the provider, and records consumption in the panel.

Panel

  • consumption by team
  • consumption by model

What it does

Token and message quotas

Each team gets its ceiling, or stays unlimited if you decide so. Enforced before the provider is ever called.

Teams and devices

Identity by credential, by IP or by an agent marker. Enabling or suspending a team does not mean touching credentials.

Model allow-list

Who may use which model, with different rules inside and outside the organisation's network.

Encrypted credentials, revocable masks

The real credential never leaves the gateway. Each client gets its own mask, revocable without rotating anything else.

Multi-provider with failover

Several providers behind one interface. If one fails, traffic continues through another without changing the client's configuration.

Admin panel

Consumption by team and by model, quotas editable live, and configuration reloaded without restarting the service.

Who it is for

Organisations already paying for artificial intelligence across several people who need to share it deliberately: know who consumes, set ceilings per team, and restrict the expensive models to whoever genuinely needs them.

Built with

PythonmitmproxyFastAPISQLiteDocker

Let's talk about your operation

Tell us which decision your business still makes by gut feel. If we can help, we'll tell you how — and if we can't, we'll tell you that too.