For developers
Build against a mock. Ship through the Control room.
Start with VidaiMock, our open-source stand-in for eight AI providers, on your laptop. When it works, send the same code through Vidai with one configuration change. Your SDK and your code stay exactly as they are.
From mock to production
# 1 · On your laptop: VidaiMock
docker run --rm -p 8100:8100 ghcr.io/vidaiuk/vidaimock:latest
OPENAI_BASE_URL=http://localhost:8100/v1
# 2 · In production: the Control room
OPENAI_BASE_URL=https://ai.your-company.internal/v1
# Your code: unchanged
client = OpenAI()
client.chat.completions.create(model="gpt-4o", ...) The same works for the Anthropic and Gemini SDKs. Or embed it in Rust tests: vidaimock = "0.3"
VidaiMock · open source, Apache 2.0
Test your AI app without paying for a single token.
Real responses, real streaming
15 endpoints, from chat and embeddings to images and moderation, with streaming in each provider's native format. It rejects malformed requests the way the provider would.
Agents and tool calls
Tool calls in each provider's own format, and agent loops that finish properly once a tool answers. Tested with Google ADK, LangGraph and LangChain.
Chaos on demand
Realistic latency, any error code in the provider's own shape, and dropped streams, set per request. Find out how retries and fallbacks behave before your users do.
Load tests that mean something
Fast enough that the mock is never the bottleneck. It's what we use to benchmark Vidai itself.
CI without keys
No provider keys in your pipeline and no flaky tests from someone else's outage.
Shaped in the open
The Rust library, for running VidaiMock inside your own tests, came from a public feature request by engineers at NVIDIA.
Everything open, everything checkable
Don't take our numbers on trust.
Run our benchmark
The benchmark script is open source. Run it on your own hardware, against VidaiMock, and compare.
Benchmark scripts →Our models, in the open
Stirling-Minimo PII: 43 MB, 77 kinds of personal data, one ONNX file. Third of 36 models on the PIIMB English leaderboard, 0.014 behind a 1.4-billion-parameter leader.†
Hugging Face →Signed releases
Every release is signed, and you can verify it before you run it. Software bill of materials and security details on request.
Trust and security →Community edition
The full Control room, free, on your own traffic. Any provider, one configuration change.
Try Community edition →Docs
Install, connect your SDKs, write your first policy, and every VidaiMock option: providers, templates, chaos and the Rust library.
docs.vidai.uk →From the blog
Why we wrote Vidai in Rust, and what Python costs on every call.
Rust vs Python for AI infrastructure →†PIIMB English leaderboard, 17 October 2026: Stirling-Minimo PII v0.1 ranked 3 of 36 at 0.904 average masking F2 across four English sets, 0.014 behind the leader, a 1.4-billion-parameter model. The leaderboard. VidaiMock figures from its README.
How we start together
Start with one team, on your own AI traffic.
- 01
Pick one app or team
Point it at the Control room. One configuration change; nothing else changes.
- 02
Watch for two weeks
Every call priced, checked and recorded. Rules start in observe mode.
- 03
See the evidence
What it cost, what it caught and what it would have stopped, from your own traffic.