Answers leave the gateway checked.
Lens sits in front of the models you already use. Every answer comes back with the sources that support it, the ones that dispute it and the passage behind every citation.
One change at the edge of your stack.
- Route to the provider you already pay for
- Bring your own model
- Prompts stay exactly as written
- Same client library
- Same request shape
- One endpoint to change
- Sources that support the answer
- Sources that dispute it, counted
- Over a billion passages behind it
- Your documents alongside the open index
- Answers cite your files like any source
- One answer across private and public sources
- Nothing you send trains a model
- Per-query audit trail
- Answers respect existing permissions
Your question runs against the full index.
Duplicates and syndication fold into one source each.
Every source lands as support, dispute or silence.
Built from what held up, with the passage behind every claim.
The finished answer is checked claim by claim before it returns.
What happens on every request.
Five steps run between your request and the response. No answer skips the check.
The gateway is one way in.
Teams pick the surface that fits how they work. Every one runs on the same platform and returns the same checked answers, so a question asked through the gateway comes back the way it would anywhere else.
- Send in an answer a model already wrote
- Get back what held up and what is disputed
- A whole set of outputs in one pass
- Compare models on the checked record
- Point your client at our endpoint
- Nothing to run on your side
Point it at work already written.
The gateway checks what your models return right now. The same check runs on answers written somewhere else, one at a time or a whole set in one pass. Either way there is nothing to stand up on your side.
Bring a workload and we will run it on your questions.