Verity
/verify · from $0.02
Verify whether a factual claim is true.
The trust layer for AI agents
Calibrated, fail-closed verification your agent pays for — per call, in USDC on Base. Verity refuses to bluff, so your agent never acts on a guess.
Live on Base mainnet · 16 services · 48 paid endpoints
Built to be wrong in the safe direction
The entire value is honesty. Three promises every service keeps:
Fail-closed
When it isn't sure, it says so and abstains — never a confident wrong answer.
Calibrated
Every response carries an honest 0–1 confidence, so your agent can set its own bar.
On-chain honest
No keys held, no accounts. Your agent pays only when it gets an answer, settled in USDC.
16 services, one trust fabric
Each is a single calibrated call. Three tiers — quick, standard, pro.
Guardrail
/check · from $0.02
Pre-action safety: allow, review, or block.
Ground
/ground · from $0.03
A sourced answer to a question.
Verify-ID
/source · from $0.02
Reputation and reliability of a source.
Citation
/cite · from $0.03
Find real sources that support a statement.
Provenance
/provenance · from $0.02
Estimate if text is AI-generated.
Sentinel
/sentinel · from $0.02
Detect prompt-injection and manipulation.
Redact
/redact · from $0.02
Find PII, secrets, and credentials.
…and Arbiter, Freshness, Sieve, Compass, Concord, Tally, Distill, and Faithful. See the full catalog.
How an agent uses it
No accounts, no API keys. Payment is the x402 standard — automatic, per call.
1. Call
Your agent POSTs to a service endpoint.
2. Pay
It gets a 402 with the price and pays in USDC — gasless, automatic.
3. Verdict
It receives a calibrated, fail-closed JSON answer.
# your agent calls a service POST verity-suite.onrender.com/check # gets 402, pays USDC automatically, then receives: { "decision": "block", "risk": 0.95, "reasons": [ … ] }