Security
Security at AgentsUse
AgentsUse security starts with limiting what the site claims and what it holds. The public catalog publishes checked facts, the AI assistance stays in draft, and the current workspace keeps its data in the visitor's own browser. This page describes those practices as they stand today.
Verified before anything publishes
Every catalog record moves through explicit states: candidate, draft, verified, then published. Candidate and draft records never get a public page. Profiles are checked against primary sources, carry a checked date, separate sourced facts from vendor claims, and leave unknown fields empty. A human reviews a record before it can become an indexable page. The full rules are on the methodology page.
AI enrichment stays in draft
Claude enrichment is built into the ingestion pipeline to classify candidates and draft verification notes and comparison points. Its output goes to the human review queue only. It never edits a public record and never publishes anything on its own, and it runs only when the feature is deliberately enabled with API credentials; the flag is off by default. A reviewer checks every note against primary sources before it can inform a page.
Workspace data stays in your browser
In the current workspace, saved stacks and the watchlist are stored in your browser on your device. They are not synced to an account, and the workspace demo key does not call a live API. Treat the workspace as a place for catalog organization, not for sensitive material: do not store API keys, passwords, or private data in it.
Examples never carry real credentials
Setup guides and configuration examples on AgentsUse use placeholders only. A published configuration is labeled config-checked when it matches the source structure, and it never contains real secret values.
Contact without exposing secrets
Contact runs through the contact page. Do not send API keys, passwords, or private data in that form; a correction only needs the page and a source link.
Report a security issue
If you find a security problem on agentsuse.com, report it through the contact page. Include the page, what you observed, and the steps to reproduce it, without including any real credentials or private data in the message.
Scope of this page
This page describes current practices only. It makes no certification claims, and it will be updated as the product changes, for example if account-synced features replace the browser-local workspace. Related reading: the privacy policy, editorial policy, and methodology.