Frequently asked questions.

How bitfence scores, what it costs, and what happens when the data isn’t there.

What happens when bitfence can’t check something?

Circuit breakers fire on known evidence only — bitfence never guesses a token is dangerous, and never guesses it’s safe. When a safety-critical check can’t run, the response carries a named degraded object saying exactly what couldn’t be verified, and the recommendation is capped at REQUIRE_HUMAN_APPROVAL. We tell you when we couldn’t check.

What recommendations can come back?

Exactly three: PROCEED (score 0–24), REQUIRE_HUMAN_APPROVAL (25–69), and BLOCK (70–100). The middle tier is deliberate — instead of guessing on an ambiguous token, the agent escalates to its human. Autonomy where the data is clear, a human in the loop where it isn’t.

Why is USDC scored differently from a new launch with the same flags?

bitfence maintains a curated trusted-token registry. Established tokens like USDC or wSOL match the registry and are assessed against their maturity; a day-old token carrying the same on-chain flags is not. The response says so explicitly — a trusted_token field appears whenever the registry matched.

Can a cached score go stale?

Every assessment carries the version of the risk policy that produced it, and cached verdicts are invalidated the moment the policy changes — no verdict outlives its policy. Cache lifetimes also scale with risk: the riskier the token, the shorter its cached score lives.

How do I pay?

Via x402 USDC micropayments, settled per query — from $0.003. Your agent’s wallet handles payment automatically as part of the HTTP exchange; no accounts, no API keys, no subscriptions. Approved integration partners can request an API-key arrangement: info@bitfence.ai.

Which chains are supported?

Solana, Base, Ethereum, Arbitrum, BSC, and HyperEVM today. Solana tokens are addressed by Base58 mint address, EVM tokens by 0x address — same endpoints, same response shape on every chain.

How fast is it?

Cached responses return in under 50ms. A cold assessment — fresh on-chain data, full scoring — runs on a sub-second path. Payment settlement on Base adds a few seconds on top; that’s the payment rail, not the scoring.