solana copy trading bots, and what to check first

solana's telegram bot ecosystem is the most developed copy-trading tooling in crypto. the bots have evolved from swap helpers into full terminals — sniping, wallet tracking, dca, mev protection, copy trading across many wallets at once. the names that come up repeatedly in 2026 comparisons are trojan, maestro, bloom, gmgn and bonkbot.

we are not going to rank them. the feature tables go stale within months, and the two things that actually determine your outcome are not on them.

the structural issue: key custody

the reason a telegram bot can execute a trade in one tap is that it holds the key. it generates a wallet for you and keeps the key material on its infrastructure, so it can sign without asking. that is the feature, and it is also the exposure.

the distinction to hold onto:

these are not two amounts of risk. they are two categories of worst case, and the difference is verifiable on-chain rather than a matter of trusting a landing page.

why this is not hypothetical

in february 2026, blockchain investigator zachxbt alleged that senior staff at axiom — a major solana trading platform — had used internal admin tools to cross-reference user ids with wallet data and front-run high-value users' trades, over roughly ten months, for more than $400,000 in profit. reporting on the allegations described business development staff with unrestricted access to live user databases, and a platform generating substantial revenue that reportedly lacked role-based access controls and audit logs that would have caught it earlier. axiom said it would conduct a full internal audit.

these are allegations rather than findings, and they concern one company. the reason to know about them is not to single anyone out — it is that they illustrate exactly what the custody model makes possible.

the seven checks, in order

1. key or allowance? the first question and the one that dominates the rest. if it needs your key, everything below is a secondary consideration.
2. does it copy exits? a bot that mirrors buys and shrugs at sells has automated the exciting half of a trade whose outcome lives in the other half.
3. flat ticket and a hard cap, independently settable? a percentage of the target's size imports their risk appetite into your account, which is the thing you were trying to avoid.
4. does it show realised slippage per trade? on thin memecoins slippage routinely exceeds the fee several times over. a tool reporting aggregate performance rather than per-trade slippage is reporting the trader's results, not yours.
5. first buys versus adds. does it distinguish a token the target has never held from a top-up of an old position? the measured effect lives in the first kind, and treating both identically dilutes it.
6. two-stage kill switch. stop new copies, and close what is open. different emergencies, and plenty of tools only have the first.
7. fee floor, not fee rate. around 1% per trade is typical. a rate with no minimum is much kinder to small tickets than a rate with one.

the thing they are genuinely good at

credit where it is due: these bots solved the hard technical problem. they watch wallets on-chain, react in about a second, and route orders competently. that is not trivial engineering and it is exactly the right place to intervene.

it matters because of how fast the edge decays. our measurements on fomo first buys found the median move peaking around fifteen minutes after the leader's fill and reverting through the hour. any tool reacting to a chain event in a second is ahead of every human in the notification queue by construction — and that gap is most of the trade.

if the trader you want to copy is on fomo

a solana bot will not help you, and this trips people up. fomo routes trades primarily on robinhood chain with several evm chains alongside, so a solana-only watcher is watching the wrong networks. worse, it fails silently — you will simply see no trades and assume the trader is quiet.

copying a fomo trader needs a watcher on the chains fomo actually uses, and it needs the trader's address, which fomo does not publish. recovering the address and what automation on fomo looks like cover that path. the custody question is identical either way, and it is the one we would still ask first.