Four steps: scope the accounts a transaction locks, read the wall, price it, send it. The monitor runs the middle two from live data.
| band | reads |
|---|---|
| p25 | the value below which a quarter of landed Jito tips fall |
| p50 | the median landed tip, in lamports |
| p75 | three quarters of landed tips sit below this |
| p95 | the top 5% tail, where next-slot sends usually need to be |
| p99 | the spike band; paying here is the expensive default |
base_fee = 5,000 lamports × signatures priority_fee = round(band_lamports × CU_limit / 200,000) jito_tip = band_lamports # optional total = base_fee + priority_fee + jito_tip
The base fee splits: half burned, half to the leader. Under SIMD-0096 the priority fee goes to the validator in full. Jito tips go to Jito tip accounts.
The public RPC answers that call with zeroes. A page built on it draws a flat line at zero and calls it congestion. Tipscope reads Jito tip_floor percentiles and names the source.
Per-account priority fees are planned for when an RPC that reports them is wired in. Until then the wall is global, and the page says so.
The feed carries its own timestamp. The status bar shows seconds since that stamp, so a frozen or cached upstream is visible rather than hidden. If the feed drops, every number switches to a labelled sample.
/api/tips.