Solutions/Apps & businesses
CoinGraph for AI wallets and wallet apps.
The most expensive mistakes happen at the signing step: a sanctioned recipient, a look-alike address, a token that can blacklist you after you buy.
When it runs: When the user, or the wallet's assistant, is about to sign a swap or a send.
Pays: Design-partner API key on a contract, plus signed webhooks for alerts.
A real answer
“Send $2K of USDC to 0x0330…e54a on Ethereum”
STOP
Recipient is on the OFAC sanctions list and flagged for a stealing attack. Swapping $5K into LINK to a Binance wallet: SAFE. PEPE: WARN, the contract can blacklist addresses and pause transfers.
How it fits
What changes with the check in place.
01
Wallet Guard checks the token and the recipient in one call: SAFE, WARN or STOP.
02
Sanctions, scam flags and contract powers (mint, pause, blacklist) read live, not from a stale list.
03
Show the reasons in the confirmation sheet; keep the proof id with the transaction.
The first step
One call. Then decide what to keep.
No account to start: x402 pays per call, MCP is free during the preview. Design partners get a key, a contract and a direct line to the founder.
curl -X POST https://coingraph.ai/api/v1/agents/wallet-guard \
-H 'content-type: application/json' \
-d '{"token":"chainlink","amount_usd":5000,"to_address":"0x28c6c06298d514db089934071355e5743bf21d60","chain":"eth"}'Other audiences
Trading agents
Right before every order, and when sizing a position.
DAO treasuries
Before a rebalancing vote, and on a schedule to catch drift.
Funds & desks
Continuously, with a push when something crosses a threshold.
Exchanges & listing
Listing review, periodic re-review, and incident response.
Compliance
Before onboarding a counterparty, and before any outbound transfer.
Research tools
Whenever a user asks about a token.
People
Every morning, and before a trade.
Other AI agents
Whenever its task touches a token or a wallet.