# Somabase

> Somabase applies your chosen policy to supported text before it reaches your model provider. Evidence collection stays a separate workflow, guided by a local agent and approved before upload.

Public pages are static HTML: no account, no JavaScript, no model call. Pages under
`https://somabase.com/app` need sign-in and are not documented here.

## Machine-readable files

- [llms.txt](https://somabase.com/llms.txt): Index of the public pages, the machine-readable files and the non-claims.
- [llms-full.txt](https://somabase.com/llms-full.txt): Plain text of every public page, from the same content source as the HTML.

## Public pages

- [What brings you to Somabase?](https://somabase.com/explore/index.html): Find your starting point: useful AI analysis, a first protected request, or a closer look at your evidence.
- [Put useful business context to work. Share fewer identifiers.](https://somabase.com/explore/business.html): Start with the problem in front of you: sensitive analysis, less evidence chasing, and clearer proof for an upcoming review.
- [A privacy policy between your app and its model provider.](https://somabase.com/explore/technical.html): Evaluate protected text completions with OpenAI, Anthropic, or OpenRouter. Keep evidence collection as a separate, explicit workflow.
- [Understand what leaves. See what the evidence proves.](https://somabase.com/explore/security.html): Evaluate two separate questions: what data a model receives, and what proof supports a control.
- [From a requirement to collected evidence](https://somabase.com/examples/database-encryption.html): Database-encryption evidence for Northstar production. This static example shows a scoped artifact summary, not a collected customer file.

## Non-claims

- Privacy API and evidence are separate products. No compliance certification is implied.
- Neither the privacy API nor artifact acceptance is HIPAA certification or automatic de-identification.