Questions, answered plainly.

What Lens is, how the checks work and how your documents are handled. If something is missing, book a demo and ask.

What is Lens?

Lens builds Lens Strata, an AI platform for organizations in any industry. It gives your teams search, chat, agents and automated work over the documents and tools you already use. Every answer is checked against its sources, every agent action is checked before it runs and a record is kept of both.

How does Lens check a claim?

For a given claim, Lens reads the sources in the Indexes you selected and marks each one: does it support the claim or contradict it. It counts them and attaches the passage behind every verdict, so you see the weight of evidence instead of one sentence. Ask whether a vendor has open enforcement actions, for example, and Lens might read two dozen sources, mark nineteen that support the claim and five that dispute it, and open each to the exact passage it came from.

What do supports and refutes mean?

Supports means the source backs the claim. Refutes means it contradicts it. Each verdict comes with the exact passage it came from, so you can act on the evidence, not a summary.

How is Smart Search different from Chat?

Smart Search returns ranked source and document results and no generated answer. You pick which Index to search across the Web, the Lens Index or your own, and you read the sources yourself. Chat generates an answer grounded in the Indexes you choose, on a model you pick, and shows the sources it used.

Do you train on my documents?

No. Documents you upload into an Index are used to answer your own queries. They are not used to train models and they are not pooled with anyone else's data.

How are my Indexes kept private?

An Index is your collection. Access follows the permissions you set, and search, chat and agents only see the Indexes you point them at. Your documents are not shared across accounts.

Which models power Chat?

Chat runs on Lens-hosted open models today and you pick the model in a selector. Bring your own model is coming, so the same checks run on a provider you already use. Smart Search has no model picker because it does not generate an answer, only an Index selector.

Can Lens still answer when the sources are thin?

Yes. When the record does not settle a question, Lens answers from the model's own knowledge and marks that answer unverified, so you get a response and know it is not backed by your sources. It does not refuse or go silent.

Are the results reproducible?

That is a design requirement, not a nice-to-have. A label that changes between reads is broken, so Lens is built so the same question over the same sources returns the same result. It is the difference between a check you can cite and a guess.

How do I get access?

Lens Strata is in early access. Request early access to join the list, or book a demo to see it on your own sources.

What can I plug Lens into?

The Lens Strata web app is the main surface. The platform also runs as a gateway in front of the models you already use, as an MCP server for the assistants you already run and as an API for teams building their own tools. The browser extension is coming.

How much does it cost?

Lens Strata is in early access and pricing is set per team as pilots come online. See the pricing page for the current tiers, or book a demo and we will scope it to your use.

Still have a question?