Floe

Live Network

Solana
Mainnet.
Right now.

Real-time data polled every 3 seconds. Success rates, TPS, priority fees — all from getRecentPerformanceSamples. Slot numbers link to Solscan for verification.

solana network — connecting...

establishing connection to mainnet-beta

tx_landing_benchmark.rs — output (peak congestion)

condition	public rpc	staked	w/ floe
Normal load	~85%	~95%	~97%
NFT mint	25–45%	~80%	~92%
Pump.fun launch	15–30%	~70%	~88%
Jito backlog	20–35%	~65%	~85%

10k tx / condition, 3 runs, median. peak congestion benchmarks.

The Problem

Your tx
gets dropped.

QUIC replaced UDP in 2022, but during congestion the slot leader's TPU port is overwhelmed. Unstaked connections = dropped first.

Your priority fee is irrelevant if the packet never arrives. It's a delivery problem.

Failure Mode

Where it
breaks.

During congestion, the slot leader's TPU port receives more QUIC connections than it can handle. Stake-weighted QoS kicks in. Unstaked connections are dropped first.

Your priority fee? Irrelevant. The packet never reaches the validator. It's not a speed problem — it's a delivery problem.

Your Transaction

signed, fee attached, ready to send

sendTransaction()

Public RPC Node

unstaked connection · no priority · shared queue

QUIC stream → leader TPU

DROPPED

failure

TPU port at capacity — connection rejected

✕Stake-weighted QoS: unstaked = lowest priority

✕TPU queue full: connection refused

✕IP rate-limited: throttled at QUIC layer

result

"Transaction expired"

retries

3/3 failed

fee

wasted

see: Solana QUIC implementation, stake-weighted QoS (SIMD-0110)

floe-relay — statusLIVE

$ floe status --network mainnet-beta

Floe Relay v0.3.2 | operational

├─ relay nodes: 12 (5 staked, 4 jito, 3 geo)

├─ avg landing: 0.41s ±0.14

├─ success rate: 88.4% (last 1h, congested)

├─ slot: 281,456,912 ↗

├─ leader: schedule-aware routing

└─ fee market: elevated

$ floe benchmark --compare public-rpc --congestion peak

public rpc22%

w/ jito64%

floe88%

Relay Status

88%
Landing
rate.

12 relay nodes across 3 regions. Multi-path delivery through Jito + staked RPCs. The same infra MEV bots use — accessible to everyone.during peak congestion. normal conditions ≈ 97%+.

How It Works

Private relay.
Multi-path.
Adaptive.

architecture — simplified

your tx

Floe Gateway<5ms

engine

Adaptive Fee

schedule

Leader-aware

selector

Route

Relay A

staked · us-east

Relay B

staked · eu-west

Jito

block engine

Leader TPU

slot 281,456,912

first path to land wins. others are discarded. zero duplicate risk.

Enter the Relay Network

Your transaction enters Floe's private overlay. Bypasses the public QUIC queue entirely. Routed through staked validator connections with priority TPU port access.

persistent QUIC streams. no handshake overhead.

Multi-path Delivery

Dispatched through 2–3 independent paths simultaneously. Jito bundles + staked RPC endpoints. Same infra the best bots use — one API call.

jito + multi-rpc. path diversity: geographic + protocol

Adaptive Fee Engine

Reads the current block's fee market in real-time. Calculates the exact minimum priority fee for 95%+ landing probability. No overpaying.

fee = f(congestion, slot_pos, account_contention)

Leader-aware routing

Reads the leader schedule and routes toward upcoming leaders' TPU ports. Wider delivery window means higher chance of landing.

getLeaderSchedule(epoch). route to slot[current+1..current+3]

SDK

6 lines to integrate.
TypeScript. MIT.

$ npm i @floe/sdkv0.3.2 · 12kb gzipped · zero deps

quickstart.ts@floe/sdk

 1import { Floe } from '@floe/sdk';
 2import { Connection, Transaction } from '@solana/web3.js';
 3  
 4const floe = new Floe({
 5  rpc: 'https://api.mainnet-beta.solana.com',
 6  // tier = wallet $FLOE balance × live price → USD tier
 7});
 8  
 9// send with multi-path routing + adaptive fee
10const sig = await floe.send(tx, {
11  priority: 'auto',        // or 'fast' | 'economy'
12  leaderAware: true,      // route to upcoming leaders
13  jitoBundle: true,       // include jito path
14});
15  
16console.log(`landed in ${sig.duration}ms via ${sig.route}`);
17// → landed in 380ms via relay-a (staked)

Network

12 relay nodes.
5 continents.

uptime99.97%

staked nodes5

jito relays4

geo-diverse3

floe-relay — node statusLIVE

node	region	type	latency	status
relay-a1	us-east	staked	12ms	ok
relay-a2	us-west	staked	18ms	ok
relay-b1	eu-west	staked	45ms	ok
relay-b2	eu-central	staked	41ms	ok
relay-c1	ap-tokyo	staked	62ms	ok
jito-01	us-east	jito	15ms	ok
jito-02	us-west	jito	22ms	ok
jito-03	eu-west	jito	48ms	ok
jito-04	ap-tokyo	jito	65ms	ok
geo-01	us-central	geo	28ms	ok
geo-02	sa-east	geo	85ms	ok
geo-03	ap-sydney	geo	92ms	ok

latency: gateway ingress → node egress. geo = path independence.

Access Tiers

Hold $FLOE. That's your key.

No staking. No burning. No lockup. USD-denominated thresholds — same dollar commitment whether you enter early or late.

·Surface

free

routes1 (public rpc)

fee enginebasic

jito—

leader-aware—

platform fee0.1%

~Drift

worth of $FLOE

routes2 + jito bundle

fee engineadaptive

jitoincluded

leader-aware—

platform fee0.05%

≋Deep

$10

worth of $FLOE

routes3 + leader-aware

fee engineaggressive

jitopriority

leader-awarenext 3 leaders

platform fee0.01%

◆Abyss

$50

worth of $FLOE

routes3 + dedicated endpoint

fee enginezero-loss

jitodedicated bundle

leader-awarenext 5 leaders

platform fee0%

api accessfull

tier resolution: on-chain balance × jupiter price oracle → USD value. connect wallet → tier instant.

Token Economics

“$FLOE is a long position
on Solana congestion.”

Demand scales with congestion

Network busy → more failed txs → more people need Floe → more $FLOE bought. Every hype cycle = buy pressure.

Low barrier, real utility

$5 gets Drift. $10 gets Deep. $50 gets Abyss. Entry is cheap, but the infra behind it isn't. Utility drives demand.

est. utilization at 5k users: 65%

Fair entry at any price

USD tiers. Whether $FLOE is $0.001 or $1, same dollar threshold. Price rises → fewer tokens needed → same access.

Subscription → position

Staked RPC = $300–1,000/mo subscription. Floe = one-time token hold starting at $5. Hold it, use it, keep it.

Bot economics

MEV bots, snipers, arb runners pay for edge. Floe: cheaper than dedicated validators, comparable landing. B2B demand floor.

demand flywheel

demand drivers:
  TPS ↑ → congestion ↑ → Floe demand ↑ → $FLOE ↑
  new protocols → more tx → repeat

supply:
  └─ DEX (tradeable)
  └─ dev (~3%)            ← vesting
  └─ community / holders

Open Source

5 crates. 14,200 lines.
All Rust + TypeScript.

cargo build — release

$ cargo build --release --workspace

Compiling floe-core v0.3.2 (crates/core)

Compiling floe-relay v0.3.2 (crates/relay)

Compiling floe-fee v0.3.2 (crates/fee)

Compiling floe-cli v0.3.2 (crates/cli)

Finished release [optimized] target(s) in 8.4s

$ cargo test --workspace

Running 247 tests

test result: ok. 247 passed; 0 failed (14.2s)

workspace modulesgithub.com/floe-north/floe

floe-corerust4,200QUIC routing, connection pooling

floe-relayrust3,800relay daemon, stake management

floe-feerust2,100adaptive fee, block analysis

floe-sdkts2,400TypeScript SDK, tier resolver

floe-clirust1,700CLI, benchmarking, diagnostics

recent commits

a3f21e8feat(relay): adaptive route scoring2h ago

8c4b1d2fix(fee): epoch boundary edge case5h ago

f7e93a1perf(core): -40% QUIC handshake overhead1d ago

2b8d4c6feat(sdk): preDeliver option2d ago

91c0fe3test(relay): multi-path failover tests3d ago

d4a72b9docs: architecture diagram update4d ago

Internals

Under the hood.

fee engine — core algorithm

 1/// Adaptive priority fee calculation
 2/// Analyzes last N blocks to predict optimal fee
 3pub fn calculate_priority_fee(
 4    recent_blocks: &[BlockFeeData],
 5    target_landing_pct: f64,  // e.g. 0.95
 6    account_keys: &[Pubkey],
 7) -> u64 {
 8    let base = percentile_fee(recent_blocks, target_landing_pct);
 9    let contention = account_contention_score(account_keys);
10    let slot_pressure = current_slot_pressure();
11    
12    // weighted combination: avoid overpaying
13    (base as f64 * (1.0 + contention * 0.3 + slot_pressure * 0.2)) as u64
14}

sliding window

Last 4 blocks → fee distribution. Targets Nth percentile per tier. Adjusts for account write-lock contention + slot leader pressure.

abyss tier

99th percentile + leader-aware routing to next 5 leaders. Significantly higher landing even during extreme congestion.

contention scoring

Tracks write-lock frequency per account. High-contention accounts (popular AMM pools) → higher fee multiplier.

Others sink.
You float.

frictionless by nature

SolanaMainnet.Right now.

Your txgets dropped.

Where itbreaks.

88%Landingrate.

Private relay.Multi-path.Adaptive.

Enter the Relay Network

Multi-path Delivery

Adaptive Fee Engine

Leader-aware routing

6 lines to integrate.TypeScript. MIT.

12 relay nodes.5 continents.

Hold $FLOE. That's your key.

Demand scales with congestion

Low barrier, real utility

Fair entry at any price

Subscription → position

Bot economics

5 crates. 14,200 lines.All Rust + TypeScript.

Under the hood.

Solana
Mainnet.
Right now.

Your tx
gets dropped.

Where it
breaks.

88%
Landing
rate.

Private relay.
Multi-path.
Adaptive.

6 lines to integrate.
TypeScript. MIT.

12 relay nodes.
5 continents.

5 crates. 14,200 lines.
All Rust + TypeScript.