Reality Check

MEV Bot GitHub

By JaredETH Research · MEV Research & Engineering · Updated 2026-05-17

Search GitHub for "mev bot" and you'll find hundreds of repositories. The most-starred public MEV bot repos include rusty-sando (Rust sandwich bot by libevm), subway-rs, sandwich-bot-py, mev-template-rs, and the official flashbots/mev-inspect-py. Most have stars in the thousands. None of them generate sustained profit in production. Production bots differ from public repos on four structural axes: strategy depth (multi-layer attacks not documented publicly), bribe ceiling (99.9% vs 80–95%), gas optimization (200–400k gas vs 600k–1.5M), and direct builder relationships that throttle untrusted bundle senders. Public repos also lack the low-latency RPC infrastructure (co-located nodes peered with Beaverbuild, Titan, rsync-builder), the MEV-Share rebate access reserved for trusted searchers, and the historical bundle data needed to tune profit thresholds. By the time a strategy lands in a public README, it's already been outcompeted on-chain.

This page covers the most popular MEV bot GitHub repos, why public code fails in production, and what changes when you license a working bot vs build from scratch.

On this page
  1. Popular MEV bot GitHub repos
  2. Why GitHub MEV bots don't generate profit
  3. The latency arms race
  4. Builder relationships matter more than code
  5. Free MEV bots vs paid MEV bots — honest comparison
  6. Open source vs purchased license

The most-starred MEV bot repos on GitHub include:

  • rusty-sando — Rust sandwich bot template by libevm. Clean architecture, good Rust patterns. Hasn't been updated for current bundle relay APIs.
  • subway-rs — fork-friendly Rust sandwich bot. Demonstrates the mempool decode → simulate → submit pipeline.
  • sandwich-bot-py — Python educational implementation. Useful for understanding the math; too slow for production.
  • mev-template-rs — generic template repo with sandwich + arbitrage skeleton.
  • flashbots/mev-inspect-py — official Flashbots indexer; not a bot, but the canonical reference for how to identify MEV transactions on-chain.

These are excellent learning resources. They are not profitable as-deployed.

Why GitHub MEV bots don't generate profit

Four structural reasons:

  1. Strategy depth. Public repos implement basic 3-leg sandwich attacks. Production bots run multi-layer sandwiches (5-layer, 7-layer), JIT liquidity, swap-JIT combos, and dual-purpose transactions. The math for these is proprietary and not documented.
  2. Bribe ceiling. Open-source bots typically cap bribes at 80–95% of expected profit. Production bots go 99%+. JaredFromSubway tips up to 99.9%. You can't outbid the market with a conservative bribe strategy.
  3. Gas optimization. Public Solidity contracts use OpenZeppelin and standard patterns — 600k–1.5M gas per sandwich. Production contracts use inline assembly and custom packing — 200k–400k gas. The gas savings alone can be the difference between profitable and unprofitable.
  4. Pool coverage. Public bots monitor 50–200 pools. Production bots monitor 2,000+, including new pools the moment they're deployed.

The latency arms race

MEV is fundamentally a latency game. The bot that decodes a pending transaction first, computes its bid first, and submits to builders first usually wins. Public GitHub bots run on generic cloud providers through public RPC endpoints — 50–200 ms latency to mempool feeds, 30–80 ms to relays. Production bots run on bare metal in low-latency data centers (Frankfurt, Zurich, Ashburn VA), peering directly with major validator nodes — sub-20 ms end-to-end. That order-of-magnitude latency advantage means production bots see opportunities first and outbid public bots every time.

Builder relationships matter more than code

Block builders aren't neutral utilities. They throttle untrusted senders, mute spammy bundle streams, and prioritize bundles from reputable operators. Established MEV operators have direct lines to builder ops teams (private comms, on-call engineers, dedicated bundle endpoints). New entrants from public GitHub forks get no such treatment; their bundles compete for inclusion at the bottom of the priority queue. This is why the same code can be profitable for one operator and unprofitable for another — the code is a small part of the system.

Free MEV bots vs paid MEV bots — honest comparison

The crypto trading bot market splits into two camps: free open-source bots (GitHub repos, hobby projects, community Discords) and paid commercial bots (licensed software, SaaS, source code purchases). Here's how the two compare on the dimensions that actually determine whether a MEV bot makes money:

DimensionFree / open-sourcePaid (JaredETH MEV bot class)
Upfront cost$05,000 USDT/yr (license) · 10,000 USDT one-time (source)
Strategy depthBasic 3-leg sandwich, simple arbMulti-layer (5/7-layer) sandwich, flash-loan arb, JIT, swap-JIT combos
Gas cost per attack600k–1.5M gas200k–400k gas (custom assembly)
Builder bribe ceiling80–95% of profit99.9% of profit
Pool coverage50–200 pools2,000+ pools (incl. new pools at deploy)
InfrastructureYou set up RPC, mempool feed, server, monitoringManaged (license) or pre-built (source)
Builder relationshipsNone — bundles compete at bottom of priority queueDirect lines to Beaverbuild, Titan, Flashbots ops
Updates & maintenanceCommunity PRs, often abandonedContinuous — strategy changes shipped to all licensees
Refund / guaranteeN/A60-day money-back guarantee
Realistic outcomeEducational; rarely profitable5,924 ETH reference deployment (JaredFromSubway)

Free MEV repos are reading material. Production MEV requires production infrastructure — either build it (months of work, six-figure capex) or license a deployment that already exists.

Open source vs purchased license

Forking a GitHub MEV bot is free but unlikely to generate profit. Licensing the JaredETH bot through JaredETH costs 5,000 USDT/yr (License to use the app) or 10,000 USDT one-time (full source code you own and self-host). The license includes the strategy depth, gas optimization, builder relationships, and capital efficiency that public repos lack. For a full breakdown of build vs buy economics, see the build a MEV bot guide.

Further reading and references

  • Flashbots mev-inspect-py Canonical reference indexer for identifying MEV transactions on-chain. Not a bot; the diagnostic tool.
  • Flashbots ecosystem repos Official Flashbots open-source organization — MEV-Boost, MEV-Share node, relay code.
  • Flashbots MEV-Inspect Open-source MEV attribution pipeline — lets you cross-check any claimed performance against on-chain reality.