For platform engineering

Stop building the plumbing around every AI call.

Teams keep their SDKs and change one setting. Translation between providers, fallback, key handling, logging and guardrails come with the Control room, instead of living in every service.

Vidai Control room · Keys
Connect your app: the address, where the key goes and the main paths, for each SDK.
Connect your app: the address, where the key goes and the main paths, for each SDK.

Your problem today

The same plumbing, in every service.

Provider keys sit in config files. Each provider wants its own SDK, so switching models means changing code. Every service carries its own retries, logging and guardrail code, and when a call fails nobody can say whether it was the provider, a limit or a policy.

Life with Vidai

One address, one key

Point the OpenAI, Anthropic or Gemini SDK at the Control room and use a Vidai key. Nothing else in the code changes, and provider keys leave your apps.

Any model through any SDK

Requests, streams and tool calls are translated both ways, and code calls model names your organisation chooses, so a new version or provider never means a code change.

MCP and tools, governed

Apps that use MCP keep using it. Their model calls go through Vidai, so every tool definition, tool call and tool result is checked by your guardrails and recorded with the call.

Refusals in words

Every refusal comes in your SDK's own format with one code to match on, and the call's record says which rule decided it, and where it was set.

Retries you don't write

Fallback chains try the next model before anything reaches your program, and the answer says what was tried.

Agents with owners

Applications and agents are registered with their own keys and owners, so a cost or a refusal lands on a system with a name, not on "a key".

What you'll use

Your answers, in the Control room.

Request logs · Why this happened

Every rule that acted, in order.

Quote the request id from any answer to find the call, and see each rule that weighed in, in the order they were applied.

Vidai Control room · Request logs
Why a call was handled the way it was: each rule, in order.
Why a call was handled the way it was: each rule, in order.

Routing · Fallback

The next model, before your code notices.

When a provider fails or rate-limits, the chain moves on to the next model, and the call record lists every model it tried.

Vidai Control room · Request logs
A call that fell back: the models tried, and the one that answered.
A call that fell back: the models tried, and the one that answered.

Models · Names

Names your code can rely on.

Apps call one name. Aliases and the model behind it can change in the Control room without a deploy.

Vidai Control room · Models
A model's name, its aliases and its id at the provider.
A model's name, its aliases and its id at the provider.

The work Vidai takes on

The platform and SRE work behind a homemade AI layer: translation across SDKs, failover, key handling, logging and the guardrail code in every service.

Related

How we start together

Start with one team, on your own AI traffic.

After a 20-minute demo, we agree a two-week pilot with you: one app or team, your traffic, your infrastructure.
  1. 01

    Pick one app or team

    Point it at the Control room. One configuration change; nothing else changes.

  2. 02

    Watch for two weeks

    Every call priced, checked and recorded. Rules start in observe mode.

  3. 03

    See the evidence

    What it cost, what it caught and what it would have stopped, from your own traffic.