Tokens flying into exchanges, picked up as they land.

Every token on the scope, ranked.

A plus sign means the labelled exchange wallets now hold more of the token than before, i.e. it flowed in. Click any line to open its holding trace, the split between exchanges and recent sampled transfers.

Each exchange, as one station.

What every exchange's known hot wallets hold across the tokens we watch, priced live off Jupiter, plus the net change since the sweep nearest to a day ago.

Individual deposits and withdrawals.

Every four minutes we pull the newest signatures from a rotating handful of hot wallets and the largest exchange token accounts of the busiest tokens, then decode a limited batch. Think of it as spot checks, not a full log. Dollar figures use the current price.

Known gaps in what this radar can see.

• N01 LABELS

  Coverage is partial. We follow a set of publicly tagged hot wallets. Exchanges operate far more, including a deposit address per customer that nobody outside can map, so plenty of traffic never crosses our list.

• N02 RESHUFFLE

  House moves show up too. If an exchange moves coins between a wallet we know and one we don't, it reads as money arriving or leaving. When both sides are tagged, the strip is marked as internal.

• N03 NET

  We see the balance, not the traffic. Zero change over an hour could still mean $50M came in and $50M went out. The strips catch individual transfers, but only a sample of them.

• N04 TIMING

  Sweeps take a few minutes. Wallets are read one after another every ten minutes, so each sweep is spread over a short span. Each window compares the sweeps nearest to 1h, 6h and 24h back.

• N05 BASELINE

  "Unusual" needs history. The score compares the past hour with the typical hourly swing we have logged ourselves. It stays blank until six hourly changes exist, and young baselines are jumpy.

• N06 USE

  Information, not a call. Deposits are often read as incoming sell pressure, yet they are equally often market makers, new listings, custody rebalancing or people staking. Nothing here tells you to buy or sell, and the site never asks you to connect a wallet or hand over a key.

How the numbers are produced.

1. Watch list

   The 80 most traded tokens on Jupiter over 24h (refreshed hourly), a fixed set of majors (USDC, USDT, PYUSD, USD1, JUP, BONK, WIF, JTO, PYTH, RAY, W, ETH, cbBTC, WBTC), native SOL, and up to 60 other positions these wallets hold worth over $250k on Jupiter's verified list. Wrapped SOL is skipped because native SOL is read directly. A token stays on the scope for 24 hours after it last qualified.

2. Sweeps

   Every 10 minutes each tagged wallet is queried with getTokenAccountsByOwner under both token programs, plus its native SOL balance, through public Solana RPC, one wallet at a time. Ten-minute sweeps are kept for 26 hours, then thinned to one per hour for 10 days.

3. Net change

   Current holdings minus holdings in the sweep nearest to 1h, 6h or 24h ago, counting only wallets that answered in both sweeps, so a failed read never masquerades as a flow. The percentage is against the earlier holdings; dollars use today's Jupiter price for both ends.

4. Unusual score

   Last-hour net change divided by the median absolute change between consecutive hourly sweeps in our own log. +8× means eight times the normal hourly swing, heading in. It appears after six hourly changes; until then the ranking falls back to last-hour dollars.

5. Strips

   Per refresh: at most 10 getSignaturesForAddress calls (20 each) across hot wallets and the top exchange token account of the busiest tokens, then at most 24 getTransaction calls. For each tagged wallet in a transaction we read its SOL change (fee added back) and token changes from pre and post balances; the counterparty is the biggest move on the other side.

6. Wallet tags

   Addresses tagged as exchanges by Jupiter's holder data, topped up with a few public explorer tags: several wallets across several exchanges. If an exchange isn't on the list, its flows simply aren't counted.