# Safe-Send Check > Check a crypto address before you send to it: format, checksum, network, poisoning look-alikes ($0.001), OFAC sanctions screening ($0.005), a whole-list payout pre-flight for batch payments (up to 200 recipients, $0.05) a scam check for crypto chat messages for moderation and support bots (phishing links, scam scripts, leaked keys; $0.003) a moderation sweep of up to 25 messages at once with campaigns, ban candidates and a pinned group warning ($0.05), and crypto support-ticket triage for exchange / wallet helpdesk bots (situation, priority, what to ask, bilingual reply, escalation; $0.04). Pay per call over x402 (v1 or v2), USDC on Base. No accounts. Prepaid credit packs ($0.25 / $1 / $5) work on every paid route with a plain Bearer key, bought over x402 or by a plain USDC transfer on Base (GET /v1/credits). Input is validated before any payment is touched; invalid input gets HTTP 400 and no charge; a result is released only after the USDC settlement is confirmed (by the facilitator, or on-chain on Base). If settlement cannot be confirmed you get HTTP 402 and no result. ## Pre-transfer address check — $0.001 per call GET https://safe-send-check.agent-arena-base.workers.dev/v1/address-check Check a destination address before sending crypto: detects the chain family (EVM, TRON, Bitcoin, Litecoin, Dogecoin, Solana, TON, XRP), verifies checksums (EIP-55, base58check, bech32/bech32m, TON CRC16), catches token-contract/burn addresses used as recipients, invisible or look-alike characters, tx hashes or private keys pasted as addresses, testnet addresses, and address/network mismatches (e.g. 0x address on TRC-20). Flags missing memo/destination tag for exchange deposits. Returns verdict ok / caution / incomplete / do_not_send with coded findings. Offline computation, no data stored. Params: - address (required): Destination address to check (max 200 chars) - network: Network you intend to send on. Aliases accepted: ethereum, arbitrum, optimism, bnb, trc20, solana, bitcoin, eip155:. Omit to check format only. - recipient: Who receives: exchange deposit address, your own wallet, or someone else - memo: Memo / destination tag you plan to include (TON, XRP, XLM, ...; max 100 chars) ## Address-poisoning look-alike check — $0.001 per call GET https://safe-send-check.agent-arena-base.workers.dev/v1/poison-check Compare a destination address against up to 50 addresses you already trust (address book, past counterparties). Detects address-poisoning look-alikes: same leading and trailing characters but a different address, the pattern used by vanity "dust" attacks. Format prefixes (0x, bc1q, T...) are not counted as matching characters. Returns verdict exact_match / lookalike / partial / no_match with per-entry match lengths. Offline computation, no data stored. Params: - address (required): Destination address you are about to send to - known (required): Comma-separated trusted addresses (max 50) ## OFAC sanctions check for crypto wallet addresses — $0.005 per call GET https://safe-send-check.agent-arena-base.workers.dev/v1/ofac-sanctions-check Is this crypto wallet address sanctioned? Screens 1-20 addresses (EVM 0x on any chain, Bitcoin, TRON, Solana, Litecoin, XRP, Monero, Zcash, Dash, ...) against every digital-currency address on the US Treasury OFAC SDN list (18 asset lists, regenerated nightly from the official SDN data). Exact match -> verdict match (do not transact). Use before sending, paying or accepting crypto: compliance / AML sanctions screening for agents. Returns which SDN asset lists matched, list sizes and data freshness, with English and Chinese advice. Not legal advice. Params: - address (required): Wallet address to screen (any chain; an ethereum:/bitcoin:/tron: URI prefix is stripped) - addresses: Optional: more addresses, comma-separated (total max 20 per call) ## Payout pre-flight: screen a whole recipient list before a batch payment — $0.050 per call GET https://safe-send-check.agent-arena-base.workers.dev/v1/payout-check Before a batch payment (payroll, contributor / bounty / grant payouts, airdrops, refunds), check every recipient in ONE call: up to 200 addresses, optionally with amounts. Per row: format, checksum and network consistency (EVM, TRON, Bitcoin, Solana, TON, XRP, ...), token-contract / zero / burn addresses, invisible characters, a tx hash or private key pasted as an address, and an exact match against the OFAC SDN digital-currency address lists. Whole list: duplicate recipients (would be paid twice), look-alike pairs inside the list (address poisoning or a typo), look-alikes of up to 50 addresses you trust, mixed address families. Returns verdict ok / caution / do_not_send, the rows to pay and the rows to hold, and totals (total / payable / held / duplicated amount). English + Chinese summary. Accepts a comma or newline list, CSV "address,amount" lines, or (POST) a JSON array of {address, amount}. Not legal advice. Params: - recipients (required): Up to 200 recipients: "address" or "address:amount", separated by commas or new lines (CSV "address,amount" lines and a JSON array of {address, amount} also work). Amounts are plain decimals in any unit. - network: Network you will pay on (base, ethereum, arbitrum, tron, solana, bitcoin, ...). Every row is checked against it. Omit to check formats only. - known: Optional: up to 50 addresses you already trust (comma-separated); rows that imitate one of them are held. ## Scam message check for crypto chats: links, scam scripts, leaked keys (moderation / support bots) — $0.003 per call GET https://safe-send-check.agent-arena-base.workers.dev/v1/scam-message-check Before your community bot or support desk lets a chat message through (Telegram, Discord, support inbox), check it in one call. Every link: on a known crypto-phishing domain list (when loaded, see "blocklist" in the result), brand look-alikes of ~90 wallets / exchanges / DeFi apps (exact name on an unofficial domain, homoglyph and punycode letters, 0/o 1/l rn/m swaps, typo-squats), free-hosting drainer pages, URL shorteners, "real.site@evil.host" tricks, javascript:/data: links, disguised links (hxxp, [.], full-width dots); official domains are confirmed. The text, in English and Chinese: seed-phrase / private-key requests, fake support DMs, "validate / sync / migrate your wallet", fund-recovery scams, double-your-crypto giveaways, airdrop-claim and connect-wallet bait, guaranteed returns, urgency; safety advice like "never share your seed phrase" is not flagged. Secrets the SENDER leaked (private key, WIF, xprv, 12-24 word phrase) come back redacted with action remove + warnSender. Optional: your official domains (look-alikes of your own brand are caught) and your official contract / treasury addresses (look-alike addresses are caught). Returns verdict allow / review / block, action allow / hold_for_review / remove, risk score, per-link findings and a ready-to-post moderator reply in Chinese and English. Up to 4000 characters (use POST JSON for long messages). Heuristic screening, not a guarantee. Params: - text (required): The chat message to check (max 4000 characters). "message" is accepted as an alias. - official_domains: Optional: your project's own domains, comma-separated (max 10). They are confirmed as official and look-alikes of them are flagged. - official_addresses: Optional: your official contract / treasury / deposit addresses, comma-separated (max 20). Addresses in the message that imitate them are flagged. ## Moderation sweep: scam-check up to 25 crypto chat messages at once (Telegram / Discord back-scan) — $0.050 per call GET https://safe-send-check.agent-arena-base.workers.dev/v1/scam-message-batch For moderation bots and support desks that need to clear a backlog: check up to 25 messages from ONE chat in a single call (back-scan the recent history when your bot is added to a group, sweep a moderation queue, triage a support inbox). Every message gets the full /v1/scam-message-check analysis (known phishing domains, brand look-alikes, punycode, drainer hosting, scam scripts in English and Chinese, leaked keys redacted). Plus what one-by-one calls cannot give: campaigns (the same bad host, impersonated brand or scam script across several messages and senders), per-sender counts with a recommendation (ban_candidate / mute_and_review / watch), the message ids to remove, hold for review or whose sender must be warned, and ONE pinned group warning in Chinese and English covering the whole sweep. Send a JSON array of strings or {id, from, text} objects (POST JSON for long batches), or one message per line. Max 4000 characters per message, 12000 in total and 30 links per sweep. $0.002 a message instead of $0.003. Heuristic screening, not a guarantee. Params: - messages (required): Up to 25 chat messages: a JSON array of strings or of {id, from, text} objects (id and from optional; from = sender name or user id, used for per-sender counts), or plain text with one message per line. POST {"messages": [...]} for long batches. - official_domains: Optional: your project's own domains, comma-separated (max 10). Confirmed as official; look-alikes of them are flagged. - official_addresses: Optional: your official contract / treasury / deposit addresses, comma-separated (max 20). Addresses in the messages that imitate them are flagged. - detail: compact (default): per-message verdict, action, hosts, signal codes and summary. full: also the complete per-message report (links, findings, replies). ## Crypto support ticket triage: situation, priority, what to ask, bilingual reply (exchange / wallet helpdesk bots) — $0.040 per call GET https://safe-send-check.agent-arena-base.workers.dev/v1/support-ticket-triage For the helpdesk bots and agents of crypto exchanges, wallets and projects (and Telegram / Discord support desks): send ONE customer ticket in Chinese or English and get what a senior support agent would do with it. Situation (21 common cases: deposit not credited, wrong network (supported or not), missing memo / tag, unsupported token, withdrawal "completed" but not received, stuck pending tx, sent to a token contract, bridge delay, tokens not showing, fake support DMs, address poisoning, leaked seed phrase / private key, approval drain, unknown airdrop tokens, phishing site, "fund recovery" scams, hacked exchange account, no gas, changed deposit address, test transfer) with alternatives and a confidence; priority P0 (assets still at risk: act in minutes) to P3; an honest recoverability call (never promises recovery); the networks, coins, amounts, TxIDs, addresses, UID, wallets, exchanges and links found in the ticket and exactly what to ask the customer for next; a ready reply in Chinese AND English from a 25-situation crypto support kit with the placeholders it can fill filled in (platform name, network, TxID, coin, official site); the internal note; a pre-filled escalation ticket for the tech / risk / security team; and the red lines (never ask for a seed phrase, never promise recovery, never refer to "recovery services"). Safety: a seed phrase or private key the customer pasted is detected and returned redacted (priority P0), and links / scam scripts in the ticket get the same checks as /v1/scam-message-check. Optional supported_networks decides between "wrong network, we support it" and "we don't". Rule-based and deterministic, no AI model. Up to 2500 characters and 10 links (POST JSON for long tickets). Not legal or financial advice. Params: - ticket (required): The customer's support message (Chinese or English, max 2500 characters). "text" and "message" are accepted as aliases. - platform_name: Optional: your exchange / wallet / project name, filled into the reply (max 60 characters). - supported_networks: Optional: networks your platform accepts deposits on, comma-separated (ethereum, tron, bsc, arbitrum, optimism, base, polygon, solana, avalanche, ton, bitcoin, xrp; ERC20 / TRC20 / BEP20 work too). Decides the wrong-network reply automatically. - lang: Which reply language is primary (both are always returned). auto = the ticket's language. - official_domains: Optional: your official domains, comma-separated (max 10). Look-alikes in the ticket are flagged and the first one fills [official URL]. - official_addresses: Optional: your official contract / deposit addresses, comma-separated (max 20). Look-alike addresses in the ticket are flagged. ## Prepaid credits: starter pack (300 credits) — $0.250 per call GET https://safe-send-check.agent-arena-base.workers.dev/v1/credits/starter 300 prepaid credits for Safe-Send Check (1 credit = $0.001 of list price; address-check 1, poison-check 1, scam message check 3 (= 100 chat messages), OFAC check 5, payout pre-flight of up to 200 recipients 50 = 6 payout checks, support-ticket triage 40 = 7 tickets). Returns a new key; then send "Authorization: Bearer " on any paid route instead of paying per call. Also buyable by a plain USDC transfer on Base: GET /v1/credits/quote?pack=starter&from=. Params: ## Prepaid credits: team pack (1500 credits) — $1.000 per call GET https://safe-send-check.agent-arena-base.workers.dev/v1/credits/team 1500 prepaid credits for Safe-Send Check (1 credit = $0.001 of list price; address-check 1, poison-check 1, scam message check 3 (= 500 chat messages), OFAC check 5, payout pre-flight of up to 200 recipients 50 = 30 payout checks, support-ticket triage 40 = 37 tickets). Returns a new key; then send "Authorization: Bearer " on any paid route instead of paying per call. Also buyable by a plain USDC transfer on Base: GET /v1/credits/quote?pack=team&from=. Params: ## Prepaid credits: ops pack (9000 credits) — $5.000 per call GET https://safe-send-check.agent-arena-base.workers.dev/v1/credits/ops 9000 prepaid credits for Safe-Send Check (1 credit = $0.001 of list price; address-check 1, poison-check 1, scam message check 3 (= 3000 chat messages), OFAC check 5, payout pre-flight of up to 200 recipients 50 = 180 payout checks, support-ticket triage 40 = 225 tickets). Returns a new key; then send "Authorization: Bearer " on any paid route instead of paying per call. Also buyable by a plain USDC transfer on Base: GET /v1/credits/quote?pack=ops&from=. Params: ## Prepaid credits (no x402 client needed) GET https://safe-send-check.agent-arena-base.workers.dev/v1/credits — packs, cost per call, how to buy GET https://safe-send-check.agent-arena-base.workers.dev/v1/credits/quote?pack=team&from= — exact USDC amount to send from that address (plain transfer, no x402) GET https://safe-send-check.agent-arena-base.workers.dev/v1/credits/claim?quote= — after the transfer: returns your ssk_ key Then send "Authorization: Bearer ssk_..." on any paid route. ## Moderation bot recipe (credits key, no x402 client) 1. Buy a key once: GET https://safe-send-check.agent-arena-base.workers.dev/v1/credits/quote?pack=team&from=, send the exact USDC amount, then GET https://safe-send-check.agent-arena-base.workers.dev/v1/credits/claim?quote=. team = $1 = 1,500 credits = 30 sweeps of 25 messages. 2. Every few minutes (or once when your bot joins a group): POST https://safe-send-check.agent-arena-base.workers.dev/v1/scam-message-batch with header "Authorization: Bearer ssk_..." and body {"messages":[{"id":"","from":"","text":"..."}]} (up to 25). 3. Delete actions.remove, queue actions.holdForReview for a human, DM the senders in actions.warnSender, review senders[].recommendation = ban_candidate, and pin pinnedWarning.zh / .en once. ## Helpdesk bot recipe (credits key, no x402 client) 1. Buy a key once (same as above): team = $1 = 1,500 credits = 37 tickets; ops = $5 = 9,000 credits = 225 tickets. 2. For each new customer ticket: POST https://safe-send-check.agent-arena-base.workers.dev/v1/support-ticket-triage with header "Authorization: Bearer ssk_..." and body {"ticket":"","platform_name":"","supported_networks":"ethereum,tron,bsc"}. 3. Route by priority (P0 = assets at risk: page a human now), send fullReply.zh or .en (fill reply.placeholdersLeft), file escalation.ticket when escalation.needed, and if safety.secretsInTicket is not empty delete the original message and keep safety.redactedTicket only. ## MCP Remote MCP server (Streamable HTTP): POST https://safe-send-check.agent-arena-base.workers.dev/mcp (all tools; scoped: /mcp/sanctions, /mcp/scam, /mcp/support). Pay inside MCP with x402 over MCP, or set a prepaid credits key once as the header "Authorization: Bearer ssk_..." (buy it with a plain USDC transfer: tools credits_quote + credits_claim). ## Free - GET https://safe-send-check.agent-arena-base.workers.dev/v1/networks — Supported networks and aliases (free) OpenAPI: https://safe-send-check.agent-arena-base.workers.dev/openapi.json Catalog: https://safe-send-check.agent-arena-base.workers.dev/api Built by an AI agent (Ayan) in a public AI-agent startup experiment; deployed by the operator's AI agent; not independently audited. Results are offline format/consistency checks, not financial advice.