Skip to content

Safety Pack for AI agents

Before an AI agent buys a token, sends money or signs something on BNB Chain, it can ask MultiScreenAI three questions: is the token safe, is this address or site a scam, and has the creator scammed before.

The checks

  • Token safety check (check_token, $0.01): Before buying a BNB Chain (BSC) token: is it a honeypot or a rug? Simulates a buy and sell, reads buy and sell tax, owner powers (mint, pause, blacklist, tax changes), holder concentration, liquidity and pool age. Returns safe, caution or danger, a 0-100 risk score and plain reasons.
  • Scam address and link check (check_scam, $0.01): Before sending funds, approving or signing: is this wallet, contract or website a known scam? Checks GoPlus reports (phishing, stealing, sanctions, money laundering, drainers) and the ScamSniffer scam lists. Returns safe, caution or danger with reasons.
  • Token creator history (check_deployer, $0.01): Who created this BNB Chain token and have they scammed before? Checks the creator wallet against GoPlus and ScamSniffer, malicious contracts and honeypots it made, how much supply it still holds and whether ownership was given up.
  • Full safety pack (check_pack, $0.02): All three checks on one BNB Chain token in one call: the token, the token contract against scam lists, and its creator. Returns the worst level across them plus each result.

Every answer has a level (safe, caution or danger), a score from 0 to 100, a one-line summary and a list of flags with plain reasons. The scoring is fixed rules, not AI, so the same token always gets the same answer. Answers are cached for 10 minutes.

Connect over MCP

Add this server to Binance Agent OS, Claude, ChatGPT, Cursor or any MCP client (Streamable HTTP):

https://www.multiscreenai.com/api/mcp

For Claude Code:

claude mcp add --transport http multiscreenai-safety https://www.multiscreenai.com/api/mcp

Or call it over REST

curl "https://www.multiscreenai.com/api/agent/v1/token?address=0x0e09fabb73bd3ade0a17ecc321fd13a19e81ce82"
curl "https://www.multiscreenai.com/api/agent/v1/scam?target=example.com"
curl "https://www.multiscreenai.com/api/agent/v1/deployer?address=0x..."
curl "https://www.multiscreenai.com/api/agent/v1/pack?address=0x..."

POST with a JSON body such as {"address":"0x..."} works too. No sign in and no API key.

Price

  • 20 free checks a day for each caller. The X-Free-Checks-Left header (or _meta in MCP) shows how many are left.
  • After that, each check is paid per call with x402 in USDT on BNB Chain, settled by Binance's B402 facilitator, so the agent needs no BNB for gas. A request without payment gets HTTP 402 with a PAYMENT-REQUIRED header; send the signed payment in PAYMENT-SIGNATURE (REST) or _meta["x402/payment"] (MCP).
  • If a check fails, nothing is charged.

Agent identity

The ERC-8004 agent card is at /.well-known/agent-registration.json. It lists the MCP endpoint and whether x402 payments are on.

Where the data comes from

If one source is down, the answer says so in sources and partial and uses the others.

Limits

These are automated checks from public data. A clean result is not a guarantee, and nothing here is financial advice. BNB Chain only for now. We keep a 10-minute cache of answers and a daily count of free checks per caller, stored as a salted hash, never as an IP address; see the Privacy Policy.