Platform

One Control room for every AI call.

People, applications and agents reach AI through Vidai with the tools they already use. Every call is checked, routed, priced and recorded as it happens, inside your own network.

Vidai Control room · Dashboard
The dashboard: what was refused, kept in region, and called by high-risk systems.
The dashboard: what was refused, kept in region, and called by high-risk systems.

How it works

The path of one call.

Every AI call your organisation makes takes the same five steps, inside your own network. Your apps change one setting; everything else happens in the gateway.
  1. 1

    Your program calls, with the SDK it already has

    OpenAI, Anthropic or Gemini, as now. The only change is the base address and the key.

  2. 2

    The key says who is calling

    Each key belongs to a person, or to an agent inside an application, in a team and an environment. A test key can't reach production models or run up production cost.

  3. 3

    Every rule that reaches the key decides

    Model access, guardrails, regions, routing and fallback, budgets and rate limits, set at the organisation, environment, team, application, agent or key. A lower level can tighten a rule, never loosen it. All of it is decided before the call reaches a model.

  4. 4

    The call goes on, with your provider keys

    To the model it may use, translated between the OpenAI, Anthropic and Gemini formats where needed. The answer comes back in your program's own format, with headers that say which model answered and the request id to quote.

  5. 5

    It is priced and recorded

    At the price in force when the call was made, charged to the team, application, agent and person behind the key. It shows in Request logs within seconds, and in cost totals and dashboards within about two minutes.

Step 1, in Python with the OpenAI SDK

# The only change: where the call goes, and the key
client = OpenAI(
    base_url="https://<your-gateway>/v1",
    api_key=os.environ["VIDAI_KEY"],
)

raw = client.chat.completions.with_raw_response.create(
    model="<a model your key may call>",
    messages=[{"role": "user", "content": "..."}],
)

# Quote this id to find the call in Request logs
print(raw.headers.get("x-vidai-request-id"))

TypeScript, the Anthropic SDK, the Gemini SDK and curl work the same way. Developers

47,200*calls a second tested on one machine, every call recorded
0.58 ms*median added to each call in that test
1configuration change for your apps

*47,200 is the highest rate tested, not a guaranteed limit. Vidai, a mock provider and the load generator on one machine, so no network time is counted: the figures are Vidai's own. 32 connections, four 15-second runs, every call recorded to PostgreSQL, guardrails and TLS off. Results on your infrastructure will differ. Method and full results

What runs where

Every part runs in your infrastructure.

Two parts you meet, the gateway and the Control room, and the database they keep their records in. A program never talks to the Control room to make an AI call, and a person never talks to the gateway.

Runs on AWS, Microsoft Azure, Google Cloud, Your data centre or Air-gapped.

For your security review

The questions an architecture board asks, answered once.

Where do prompts and answers go?
Through the gateway to the provider you chose, and nowhere else. Whether full content is kept depends on your logging setting; reading it is a role of its own, and every read is recorded.
Where are the provider keys?
Held by Vidai and used by the gateway. They never sit in your apps, and the Control room shows them by label, never by secret.
Where are content checks run?
In the gateway, with patterns, and on VidaiGuard beside it. The data being checked stays in your infrastructure.
Does anything call home?
No telemetry, and the licence is checked offline. Our public rate-card server for model prices is optional; host your own (Enterprise) or enter prices by hand. Only price lists are fetched, never call data.
How fast does a change take effect?
Within seconds on every gateway node. System › Status shows each node and whether it has the latest change, and Stop AI traffic is read back from every node, so a partial stop never reads as a full one.
What about high availability?
Run several gateway nodes behind your load balancer, with autoscaling. Enterprise customers get a Helm chart and Terraform, and support with an SLA.

Region rules, VidaiGuard, signed evidence and events to your SIEM come with Enterprise. Trust and security · Beyond the gateway

Editions

Start free. Grow into the whole company.

Community is a complete governance platform on its own. Enterprise adds single sign-on, VidaiGuard, region rules, signed evidence and events to your SIEM. A licence problem never becomes an outage.

Compare editions →
Vidai Control room · Settings
Your edition, its limits in use, and what it includes.
Your edition, its limits in use, and what it includes.

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.