Conge

Blockscout API v2 is the primary source for almost everything — whether an address is a contract, the token's name, symbol and supply, its holder list and holder count, whether its source is verified (and its ABI), whether it's an upgradeable proxy, when it was deployed, and market data. It's a plain read-only HTTP API; Conge never needs a wallet or a signature to query it.

Results appear progressively. As soon as the token and holder lookups answer, the result card shows the token's name, supply, holder concentration, holder count and market data. Contract verification, owner privileges and token age each show "Checking…" and fill in on their own as their lookups return — under load, some explorer lookups take several seconds. Every Blockscout request gets 10 seconds; one that doesn't answer in time shows as Unknown, with the reason, instead of holding up the rest.

A small Cloudflare Worker acts as a read-only RPC relay for the two checks Blockscout has no field for — reading a contract's owner(), and the sell simulation — plus a lightweight connection check (the chain ID) shown under Network status. It exists because the chain's public RPC endpoint has been unreachable from some networks and devices (a TLS certificate error) and doesn't send the CORS headers a browser needs. That endpoint is also shared by everyone on the chain and tightly rate-limited, so the Worker retries a rate-limited call a few times with a short backoff rather than failing straight away.

The Worker only forwards a fixed allow-list of read-only calls (eth_chainId, eth_blockNumber, eth_call, eth_getCode), and only for requests coming from this site — nothing that could move funds, change state, or sign anything passes through it. It's also optional: if it's ever unreachable, the owner and sell-simulation checks show Unknown and everything else still works.

Conge only scores tokens. Before running any check, it confirms there's a token at the address, and stops with a plain explanation — rather than a risk card full of empty fields — when there isn't:

• A wallet address (no contract code) — there's nothing to scan.
• A contract that isn't a standard token — a router, a pool, or any contract with no token name, symbol, or decimals. Conge says it doesn't look like a standard token and can't analyze it.
• An address the explorer doesn't list as a token and can't otherwise identify right now — Conge says it couldn't confirm what the address is, and suggests trying again.

If the explorer's token lookup itself fails or times out, that proves nothing either way, so the scan goes ahead and each affected check honestly shows Unknown.

Every check is transparent and rule-based — no model, no hidden scoring. These are the exact thresholds Conge uses.

The overall level is simply the worst result among the checks that actually produced data. Conge never uses the words "safe" or "secure".

• High — at least one check came back High.
• Medium — no High, but at least one Medium.
• Low — no Medium or High findings in these automated checks. There may still be minor (Low) notes, and Low is not a guarantee.
• Insufficient data — none of the checks that count produced usable data, so there's no honest verdict to give.

A check that couldn't produce a result is shown as Unknown, grouped under "Info / unknown". Unknown checks never count toward the level — not as a red flag, and never as a pass. Market data and the owner check are informational and don't raise the level either.

While checks are still running, the gauge reads "Finalizing risk level…" and the needle sweeps; it only settles once every check that affects the level has finished.

None of this is an audit or a guarantee — it's a fast, transparent read of public data. Specific, known gaps:

• Buy/sell tax. The sell simulation confirms a transfer to a pool isn't outright blocked, but can't measure a fee taken on top of it.
• Router-only logic. Some tokens only misbehave when the caller is a specific DEX router mid-swap — a direct transfer simulation from a holder doesn't reproduce that.
• Time- or amount-based traps. The simulation moves 1% of a holder's balance, right now. A token that only blocks larger transfers, or only after or before a certain time, can pass every check here while still being a trap in practice.
• Pools outside the top holders. The sell simulation only looks for pools among the top 5 holders, and only recognizes pools that answer token0()/token1(). When it finds none, it says so (Unknown) rather than passing.
• Name-based checks aren't a bytecode audit. The owner-privilege check looks for functions by name in a verified ABI — a differently-named function with the same effect won't be caught.
• A scan sometimes can't finish every check. The owner and sell-simulation checks depend on a shared public RPC that rate-limits under load, and some block-explorer lookups can be slow or time out. Affected checks show Unknown with the reason — see the FAQ.

Always verify independently on the block explorer before acting on anything Conge shows you.