Raven doesn't just scan. It issues signed evidence receipts that state their own scope and declare what they miss.
Signed, scope-bounded evidence receipts for agent commerce. Every receipt is derived from decoded on-chain bytes, states exactly which checks ran and which did not, declares its coverage gaps, and is verifiable offline against a signer key you pin yourself. When the evidence can't be established, Raven refuses and issues no receipt — it never guesses. A receipt is evidence, not a decision, and never an authorization to act.
Name a token. We'll return a signed Raven evidence receipt.
Three rules most scanners break
Risk scores invite false confidence. Raven publishes findings, not vibes.
Fail closed
If an account can't be fetched, a layout can't be decoded, or an extension isn't recognized, that becomes an explicit finding or coverage gap — never a silent all-clear. An unreadable token is reported as not verifiable, not "probably fine".
Signed & replayable
Every receipt ships with a domain-separated ed25519 signature over a canonical payload hash. Anyone can verify what Raven observed, about which mint, at which slot — offline, forever, against a key they pin themselves. Signature validity, key trust, and freshness stay three separate answers.
Coverage gaps declared
Every receipt lists what was NOT evaluated (coverageGaps) alongside what was. Anything listed there is unverified — silence is never safety. New evidence never silently rewrites history: engine and rules versions are recorded, and changes are segmented from coverage expansion.
Archived signed evidence from the verifier
Legacy v2 captured examples. Retained for historical comparison; these cards are not the current receipt-v1 contract and not a live trust/status viewer. New integrations use POST /receipt/v1.
These are archived compact receipt views derived from signed hosted-verifier responses, not a live status feed. Every card shows its issued date, observed slot, and current age so stale proof cannot masquerade as fresh evidence. Note USDC: Raven recorded active issuer authorities as triggered findings at issuance.
Archived receipt views could not be displayed. No freshness claim is being made.
Raven verifies on-chain evidence. It does not predict price or guarantee safety.
Verify a receipt yourself
Trust comes from your own pin. Each receipt v1 signs a canonical payload hash under the raven-receipt / v1 domain. Use the ratified-but-unpublished raven-receipt-verifier package name with the primary receipt-v1 test vector.
Three ways to call Raven
MCP — for Claude & agent frameworks
One tool: raven_verify_token. This local developer path is unsigned; use the hosted receipt-v1 API when you need a signed customer deliverable.
npx raven-verify-mcp
Hosted API — signed evidence receipts
POST a mint, get a signed deliverable. Server-side RPC; callers never supply infrastructure.
POST /receipt/v1
x-api-key: <your key>
{"mintAddress":"...",
"tokenProgramAddress":"..."}
Primary customer path. Alpha — keys by request. POST /verify is legacy v2 compatibility only.
ACP — agent-to-agent commerce
RAVEN's Virtuals ACP integration is in sandbox and is not publicly listed. Sandbox agents are not discoverable in the ACP marketplace and cannot be hired from the public registry.
offering: solana_verify_token
Not graduated. Raven claims no public validation by the Virtuals ecosystem, and no live escrowed ACP jobs.
Request access
Alpha keys are issued by hand while we keep the quality bar high. Tell us what you're building and the call volume you expect. Beta holder-concentration evidence (venue-adjusted: pump.fun curve and PumpSwap vaults excluded with on-chain proof) is available per key.