Jito Bundles Explained: Tips, Atomic Execution and MEV on Solana

A Jito bundle is a list of up to five Solana transactions that run in order and atomically: either every transaction lands in the same block, or none does. The sender attaches a tip, and the Jito block engine runs auctions to decide which bundles validators include. Solana MEV bots use bundles to avoid partial, loss-making fills.

Why bundles exist

Without bundles, a multi-step trade can break halfway. Imagine buying a token on one pool and selling it on another. If the second leg fails after the first succeeds, you hold a token you did not want at a price that has moved. Bundles remove that risk by making the sequence all-or-nothing. They also give searchers a defined way to pay for priority, instead of guessing at network-wide fee levels.

The documentation for the system lives at Jito’s docs, and the commercial vendors that build on it summarise it in their own words. CryptoMEV Bot’s article on Jito MEV on Solana describes the client, the block engine and an optional relayer. Its figures on volumes, yields and revenue are not sourced, so check anything numeric against Jito’s own materials before using it.

The moving parts

Component	Role	Why a bot cares
Jito validator client	Modified validator software that accepts bundles	Only slots led by such validators can include bundles
Block engine	Receives bundles, simulates them and runs auctions	The place you submit to, and where tips compete
Relayer	Optional proxy in front of validator ports that filters spam	Affects how transactions reach the leader
Searcher	You or your bot	Builds the bundle, picks the tip
Tip account	On-chain account that receives tips	Your bundle must include a transfer to it

How a bundle is built

1. Find an opportunity. Your strategy spots a price gap, a liquidatable position or a post-swap price move.
2. Build the transactions. Up to five, ordered. One of them includes a tip transfer in lamports to a Jito tip account.
3. Simulate. Run the sequence against current state. If it would fail or lose money, do not send.
4. Submit to the block engine. The engine simulates again and enters the bundle into the auction.
5. Wait for inclusion. If your bundle wins and the leader is running the Jito client, it lands atomically. If not, it is dropped.

Solana’s account locking explains why bundles compete directly: two bundles that write to the same accounts cannot both execute in the same position, so higher tips tend to win. You can read how base and priority fees work in the Solana fee documentation; tips are a separate mechanism layered on top.

Setting a tip without destroying your margin

Tip strategy is the real skill. A common framework is to tip a share of your estimated profit and adjust with experience:

• Estimate net profit after trading fees and slippage.
• Pick a tip share. A high share wins more often but leaves less per win. A low share loses more often.
• Track win rate and realised profit per strategy, and tune the share where total profit peaks, not where win rate peaks.
• Cap the tip so a bad estimate cannot turn a win into a loss.

Key takeaway: a bundle that wins every auction is a bundle that is tipping too much. Measure profit per attempt, not just how often you land.

For ordinary users the number is far smaller. Some wallets let you attach a modest tip to get priority or protection, which is relevant to avoiding sandwich attacks.

What bundles do not do

Bundles do not guarantee profit, and they do not remove competition. Another searcher can see the same opportunity and tip more. Prices can move between simulation and inclusion. A leader that is not running the Jito client cannot include your bundle at all, so timing around leader schedules matters. Most searchers also run simulations constantly and discard far more candidates than they send, as a matter of cost control.

Bundles also do not make harmful strategies harmless. The same mechanism that lets arbitrageurs work cleanly lets sandwich bots wrap a victim’s trade, which is why sandwich attacks on Solana remain contested. The vendor page for a Solana MEV bot describes Jito bundle execution as private, atomic transactions to validators; it does not say what tip levels the software uses or how it chooses them.

A worked example with made-up numbers

The numbers below are invented to show the arithmetic, not to suggest what real trades earn. Suppose your simulation shows a two-pool arbitrage that returns 0.040 SOL of gross edge on a trade. Trading fees on both pools and expected slippage take 0.015 SOL, leaving 0.025 SOL. Base and priority fees are a small fraction of that. If you tip 40 percent of the remainder, the tip is 0.010 SOL and your net on a win is 0.015 SOL.

Now add the failure side. If you send ten such bundles and only three land, you collected three wins and paid for the infrastructure behind ten attempts. Whether that is profitable depends on server, RPC and data costs, which are fixed regardless of how many bundles win. This is why professionals track profit per day after all costs rather than celebrating individual wins.

Tip share	Win rate (illustrative)	Net per win	Verdict
10%	Low	High	Lands too rarely to matter
40%	Moderate	Moderate	Often near the sweet spot
80%	High	Very low	Wins the auction, gives away the edge

Common mistakes with bundles

• Skipping simulation. Sending without checking current state wastes effort and invites losses.
• Ignoring leader schedule. A bundle sent when the leader is not running the Jito client cannot land in that slot.
• Hard-coding tips. A fixed tip fails whenever competition changes. Make it a function of expected profit.
• Trusting one number. Dashboards that show only win rate hide whether you are making money.
• Forgetting the tip transfer. The bundle must include it, or it will not enter the auction.

If you want to see how the build process is framed from a vendor’s perspective, CryptoMEV Bot’s guide to making an MEV bot is one example; read it alongside primary documentation, not instead of it.

What to read next

If you are evaluating software rather than building, ask how a bot sets tips, how it simulates, and what it does on failure, and compare the answers with our criteria for judging a Solana MEV bot. If you plan to write your own, how to build a Solana MEV bot walks through the order of work. Features and fee schedules change, so confirm current behaviour in the official documentation before you rely on any number in this article.