Every VerityLayer check
Independent, fail-closed verification your AI agent can call before it acts — each pay-per-call via x402, each returning an Ed25519-signed receipt you can verify yourself, free.
Action guardrail for AI agents
An independent allow / review / block before your agent does something irreversible.
Grounded answers with sources
A sourced answer to a question — with citations and an honest confidence, or an abstention.
Source trust scoring
Score the reputation and reliability of a source before your agent trusts it.
Citation verification
Find real retrieved sources that actually support a claim — and catch fabricated ones.
AI-text detection
Estimate whether a passage of text was AI-generated.
Freshness & staleness check
Check whether a claim is still current as of a given date.
Claim adjudication
Judge which of several conflicting claims the evidence actually supports.
Prompt-injection detection
Detect injection and manipulation in untrusted text or tool output before you act on it.
PII & secret detection
Flag PII, secrets, and credentials before text leaves a trust boundary.
Content moderation
Screen content against a publishing policy before your agent posts it.
Decision support
Pick the best option under constraints, judged against evidence.
Policy compliance
Check text against a policy, citing the exact clauses it violates.
Math & figure check
Sanity-check the math and figures inside a piece of text.
Grounded fact extraction
Extract only the facts literally present in the text — nothing invented.
Faithfulness check
Check a summary or translation for faithfulness to its source.