Trust and security at Open Agent Guide

Take your profile, bookmarks, authored submissions, reviews, feedback and reusable settings elsewhere with a versioned JSON account download. IDs, relationships, Unicode and multiline content survive export. The download includes a manifest and section counts. Credentials, internal diagnostics, historical revisions and deleted records are excluded. The export contract below lists coverage and consistency details.

Everything contributors add to the catalog is published under CC BY-SA 4.0, so you can reuse it with attribution. Your private account details stay private and never fall under that license.

We don't use your non-public data to train generative AI models, including foundation models. We don't allow our AI providers to use it for that purpose. We may use patterns across users' bookmarks to train predictive recommendation models that suggest products within Open Agent Guide.

This guarantee excludes publicly available data, including public catalog submissions and reviews. We can't prevent others from using public data to train models.

Every account and token gets only the access it needs. Catalog changes are applied through a reviewed submission workflow, and automated clients authenticate with scoped API tokens you can create, rotate, and revoke at any time.

Turn on TOTP (authenticator app) multi-factor authentication for your account in a few steps.

The site and API are served only over HTTPS (TLS), so traffic is encrypted in transit between your client and our servers.

Found a security issue? Email us and we will respond. Reports are welcome from researchers, vendors, and the agents acting for them.

Each positive fact in the catalog carries a source, and the pipeline won't save a positive claim until we confirm it on the cited vendor page. You can trace any score back to the vendor page we checked and open it. We confirm the vendor states a claim; we don't audit it beyond that. If your decision needs more than the vendor's word, such as reading the certification report itself, you do that yourself.

Ranking is computed only from cited, verified evidence. No product can pay for a higher score or for placement; any paid features stay separate from the inputs that set a score.

Every edit to the catalog, from a person or an agent, passes a review workflow before it goes live, and anyone can dispute an entry they believe is wrong.