RAVEN. / request access

Request access

Tell us what you're building and name one token. We will run it through the live verifier and reply with a sample signed receipt — findings, coverage gaps, verdict, signature — plus an API key if it is a fit.

Paste whatever you have: the token's mint address, its name or ticker, or just a DexScreener/Birdeye link. We'll find it and run the check.
What we should call you in the reply. First name or a handle is fine.
Where we send your signed receipt. Never shared, never published — and not included if you submit via GitHub.
One line about what you're building or doing — a trading bot, an AI agent, a wallet, a community you moderate. "Just checking a token before I buy" is a perfectly good answer.
Helps us reply at the right level — beginner answers get plain language, developer answers get curl and config.
Where the receipt fits in your flow. Examples: "check a token before I ape", "my agent verifies every token before trading", "screen launches for my group".
The yes/no moment where the evidence matters. This is the single most useful thing you can tell us.
Missing evidence you care about most Raven lists what it has NOT checked in every receipt. Which gap would you most want closed?
Raven works without any storage — receipts are yours to keep wherever you like. This just tells us whether to prioritize storage guidance and adapters. See /receipt-storage.html.
How you'd rather call Raven. Hosted = we run it, you get signed receipts by key. ACP = your agent pays 0.1 USDC per job, no key. MCP = run it locally for free. "Unsure" is fine — we'll recommend one.
A rough guess at how many token checks you'd run per day once you have a key. It only helps us size your rate limit — guessing low or high is fine.
An advanced extra check: how concentrated the token's holders are — with known pool and venue wallets factored out, so a liquidity pool doesn't get mistaken for one giant whale.

GitHub submissions are public — use email if sharing contact details. Either way: we'll run your token and reply with a signed receipt.

Raven Scout — how to read what you'll get back

The engine outcome will be one of: risk (a rug-capable control exists), warning (something needs eyes, or important surfaces are unverified), pass with info (clean on checked surfaces, notes attached), or unknowable (the chain couldn't be read — we don't guess).

When something cannot be checked, the receipt says why: the account could not be fetched, a layout could not be decoded, an extension is unrecognized, or the signer is unavailable — each becomes an explicit finding, gap, or unknowable, never a silent pass.

Raven never says "safe". If coverage gaps remain, the honest answer is: not enough evidence for a full pass — and the receipt will list exactly which gaps those are.