SuperAdmin
Central vault for provider keys and model assignments
The single place where all third-party AI keys live and where each microservice is told which model to use. Change a provider or a model once — the whole platform switches over instantly.
- 1 Vault for all provider keys
- Live Model swaps propagate instantly
- 100% AI calls go through the shared client
SuperAdmin solves a quiet but critical problem: where do you keep your API keys, and which model runs in which feature? Spread across configs it becomes brittle and insecure. Centralised, it becomes one clean switchboard.
Every provider the platform talks to — OpenAI, Gemini, ElevenLabs, and others — has its keys stored here, entered once by the platform admin. Every feature that calls a model — the chat builder, the agent builder, the knowledge base, the site crawler, the scenario engine — picks its model from a dropdown bound to the same registry.
You can point heavy work at a stronger model and light work at a cheaper one. A switch here propagates instantly to every service, with no redeploy and no per-feature fiddling. If a key stops working, you rotate it in one screen and every dependent service picks up the new credential on the next call.
The panel is deliberately spare: a list of services on the left, a model and status column on the right. No decorative chrome. Access is locked to a designated platform admin — regular users cannot see SuperAdmin exists.
Usage is tracked per service: requests, tokens, rough cost. That is enough to answer the questions that actually matter — which feature is eating the budget, where is a provider throttling us, did that traffic spike translate into spend?
SuperAdmin also offers a shared outbound channel to each provider. Instead of every microservice writing its own HTTP client, retries, and rate-limit handling, they all call through one library that inherits keys, model choice, and telemetry. That is what keeps the platform consistent and predictable as new features get added.
The point is boring in the best possible way: key rotation should be a five-minute job, not a three-team ticket. Model migration should be a dropdown change, not a rewrite. SuperAdmin makes both of those true.
Everything SuperAdmin handles for you
-
Unified key vault
All provider credentials kept in one protected store, rotated from a single screen.
-
Per-service model map
Every feature declares which model it wants; admins pick the actual model from a controlled list.
-
Instant platform-wide swap
Change a model or a key once — every dependent service picks it up on the next call.
-
Usage telemetry
Requests, tokens, and approximate cost broken down per service, so spend stays visible.
-
Strict access
Only a designated platform admin can see SuperAdmin; the surface is invisible to everyone else.
-
Shared AI client
One outbound library used by every feature, with retries, rate-limit handling, and logs built in.
Every AI-powered service reads from SuperAdmin
SuperAdmin is the dependency that nobody sees but everyone uses. Any feature that calls a model asks SuperAdmin for the key and the current model assignment, then talks to the provider through the shared client.
- ChatBuilder picks the chat-tier model from the registry, including voice and avatar variants.
- AgentBuilder offers only the approved model list when composing an agent.
- PromtBuilder uses the rewrite-tier model for prompt drafts and variants.
- KnowledgeBase runs the ingest pipeline against the model marked for research tasks.
- SiteCrawler calls the extraction model whenever it parses a page.
- AIFlow evaluates conversation triggers against the model assigned to scenario matching.
- AIM CRM uses the registry when the agent-in-chat writes or reads customer records.
Wire SuperAdmin into your product today
Book a consultation with our founders and we'll walk you through the whole microservice stack — not just this one — live on your domain.