Four things that share one name
The disambiguation done properly. Each of the four categories is traced through the same three questions - who holds the keys, where the money moves, and what evidence survives afterwards - because those three answers stay stable even when the vocabulary keeps changing.
The one-sentence version
Launch tooling, market-activity tooling, execution bots and coordinated buying groups are four separate categories that a single careless phrase has merged into one.
The meaning, stated four times
The pump bot meaning problem is not that the phrase is technical. It is that it is empty. The phrase carries no information about who operates the thing, who pays for it, or what happens if nobody else shows up, and those are the three facts that decide whether you are looking at a product or an arrangement. So here are four separate definitions, each written as though the other three did not exist.
Launch tooling. A program that a token team runs in the first seconds of a token becoming tradable, whose job is to place a specific set of transactions into the earliest blocks in a particular order. The job is narrow, it is over quickly, and it either happened or it did not.
Market-activity tooling. A program that sends a schedule of small buys and sells on one token from wallets under its own control, for the purpose of producing recorded trading over a period. The output is a record, not a position.
Execution automation. A program that trades according to rules on behalf of whoever owns it. Snipers, arbitrage bots, grid traders and rebalancers all sit here. The operator carries the profit and the loss, and nobody joins.
Coordinated buying arrangement. A group of people who agree to buy the same thinly traded asset at roughly the same time. There is no deliverable. The mechanism is the participation itself, and the price movement it produces is what later arrivals buy into.
Four definitions, one phrase. The rest of this page runs each of them through the same three questions, because the answers stay stable even as the marketing language changes underneath them.
Who holds the keys
On a public chain, a wallet is a keypair, and whoever holds the private key controls the funds in it. That fact turns custody into the first useful axis, because the four categories answer it very differently.
In launch tooling, the token team typically holds every key involved. The program is something the team runs, and the funds passing through it are the team funds. A vendor may supply the software without ever touching the keys, which is the arrangement most teams should prefer.
In market-activity tooling, the program generates and funds a set of worker wallets. Who holds those keys is the central commercial question about the whole category, and the honest answer varies by vendor: some hand the keys to the customer, some keep them, and some never make it clear. It is worth noticing that the words on the marketing page rarely settle this. The keypair model itself is plainly documented in the public Solana program documentation, so there is no technical excuse for vagueness.
In execution automation, the operator holds the keys and nobody else has a claim on them. If a stranger offers to run automation using your keys, the arrangement being proposed is custody, and it should be evaluated as custody rather than as software.
In a coordinated buying arrangement, everyone holds their own keys, which sounds reassuring and is not. Nobody is taking your key. The loss mechanism does not need your key, because you will place the losing trade yourself, voluntarily, in your own wallet.
Following the money
Custody tells you who can move funds. Following the money tells you who ends up with them. Writing the flows out one line at a time makes the fourth category stand out immediately.
The first three lines all end somewhere ordinary: a vendor, a validator, a pool, or the operator own balance sheet. Value is spent on a service or risked on a position, and the total quantity of value in the system is not being redistributed between the people reading the same chat message.
The fourth line ends somewhere different. No service was rendered and no new value was created, so every unit that one participant gains came from another participant. This is the property that makes the fourth category different in kind rather than in degree, and it is why no improvement in execution, fairness rules or tooling changes the outcome for the group as a whole.
It is worth sitting with that for a moment, because it is easy to read quickly and file away. A category in which participants are paid out of the contributions of later participants does not become safer when it is well organised. It becomes more efficient at doing the thing that harms the people at the back.
What each one leaves behind
The third axis is evidence. Everything that touches the chain leaves transactions, but the four categories differ sharply in how much those transactions can tell you afterwards.
Launch tooling leaves a checkable moment
Because the job is defined by block position, you can look up the token first blocks and see whether the transactions landed together and in order. Either they did or they did not. This is the most falsifiable of the four claims, which is one reason vendors in this niche tend to be specific.
Activity tooling leaves a countable period
You can count swaps, sum their sizes and check the time distribution. If a vendor promised a quantity of trading over a window, that promise is arithmetic against public data. The wallets involved are usually recognisable as a cluster, because funding patterns are visible.
Execution automation leaves a private record
The trades are public but the intent is not. From outside, a well-run arbitrage program and a badly run one look similar until you reconstruct the accounting, which only the operator can do properly.
A coordinated arrangement leaves an ambiguous burst
You can see a concentration of buying followed by selling. What you cannot see is who was in which position or who knew what and when. Anyone can open a token address on a public Solana explorer and observe the shape, and the shape alone will never prove intent. That ambiguity is not a bug from the operator point of view.
The four on three axes
Everything above, compressed. The middle column is the one that most often gets misread, because it is the one marketing language works hardest on.
| Category | Key custody | Where value ends up | Evidence quality |
|---|---|---|---|
| Launch tooling | Team, usually; vendor may never touch keys | Vendor fee plus network fees | High: block position is checkable |
| Market-activity tooling | Varies by vendor; the question to ask first | Vendor fee, venue fees, validators | High: swaps are countable |
| Execution automation | Operator, exclusively | Operator profit or loss | Medium: public trades, private accounting |
| Coordinated buying | Each participant holds their own | From later participants to earlier ones | Low: a visible burst, an invisible order |
Five sentences that hide the difference
These are constructions rather than quotes, and each is worth learning as a shape. None of them is necessarily dishonest; each is simply ambiguous in a direction that costs the listener money.
- "We use a bot to build the chart." This is true of activity tooling and true of a coordinated arrangement, and the sentence does not say which. Ask who owns the wallets doing the buying and who is paying for the trades. Two questions, and the ambiguity collapses.
- "Get in before the bot starts." Tooling has no before that matters to an outsider, because you are not a customer of it. A sentence that offers you an early position is describing a sequence of buyers, which is the fourth category wearing the vocabulary of the second.
- "It is automated, so it is neutral." Automation describes how instructions are sent. It says nothing about who benefits. A program can execute a perfectly neutral schedule on behalf of an arrangement that is not neutral at all.
- "Everyone wins if we all hold." This describes a coordination problem in which defecting is profitable and being last is expensive. When a plan requires the participants to act against their own interest for the plan to work, the plan is not a plan, it is a hope with a schedule attached.
- "The volume proves people are interested." Recorded activity proves that trades happened. It does not distinguish trades placed by people forming an opinion from trades placed by a program working through a schedule. Depth, holder distribution and the persistence of activity after the campaign ends carry far more information.
Where the safety questions differ
Because the four categories fail in different ways, the questions worth asking about them are not interchangeable.
For launch tooling, the questions are technical and narrow: what exactly does it do to the transaction ordering, what happens when the attempt fails, and does the team retain control of its own funds throughout. The failure mode is wasted spend and a messy launch, not a transfer of wealth to a stranger.
For market-activity tooling, the questions are commercial and custodial. Who generates the worker wallets and who can move funds out of them. What happens to the residual balances at the end of a run. Whether the quoted price covers venue costs or only the vendor own margin. Whether the promised quantity is stated in a way that can be counted afterwards. This is exactly the ground that the question is a Solana volume bot safe is meant to cover, and a vendor who answers it with adjectives rather than mechanics has told you something useful anyway.
For execution automation, the questions are about your own discipline: what the rules do in conditions you did not anticipate, how much of your capital any single rule can commit, and whether you would still be comfortable with the position if the program stopped running halfway through.
For a coordinated arrangement, there is no equivalent list, because the failure is structural rather than operational. You cannot ask a better question and arrive at a version of the arrangement that is safe for you, since your safety would have to come out of somebody else pocket in the same group.
How the language drifts
Vocabulary in this market moves fast, and it moves in a predictable direction: away from words that describe structure and towards words that describe feeling. A term that starts as a precise description of a mechanism gets adopted as a general compliment, loses its edges, and eventually needs replacing by a new precise term, which then repeats the cycle.
You can watch this happen to almost any term here. Words that once meant a specific technical arrangement end up meaning roughly "good" or roughly "coordinated", and at that point they stop being useful for telling anything apart. The defence is not to memorise the current vocabulary but to keep asking the three structural questions, because custody, money flow and evidence do not drift.
This is also why a vendor being precise is itself a signal. Precision costs a seller nothing when the product is real, and costs an arrangement everything, because the arrangement only functions while the participants keep their own position in the sequence vague in their minds. When somebody responds to a structural question with more enthusiasm, that is data.
Choosing between real tools in the second category is its own separate exercise, and it turns on the same structural properties described here rather than on the language. A professional Solana volume bot is distinguishable from an amateur one mainly by what it will tell you before you pay, which is a habit of mind you can carry into any of these categories.
Putting it into practice
Take whatever you were told, and rewrite it in the shape of the three axes. Who holds the keys. Where does the money end up. What could be checked afterwards, and by whom.
Most of the time this exercise takes under a minute and produces a clear answer. When it does not produce a clear answer, that is itself the answer: a category with customers can always be described this way, because the description is what the customer is buying. A category with participants resists the description, because writing it down plainly would make the position of the person at the back of the queue too easy to see.
If the thing in front of you turns out to be the fourth category, the pages under how people get hurt cover what happens next. How coordinated buying works describes the mechanism at the depth needed to recognise it, and why late buyers lose works through the arithmetic that decides the outcome. If you want the compressed version of this whole page as something you can run against a specific offer, the tool-or-scheme checklist is built for exactly that.
Questions readers send about this page
What does pump bot mean, in one sentence?
It means whichever of four unrelated things the speaker had in mind: launch tooling, market-activity tooling, general execution automation, or a coordinated buying arrangement in a chat group. The phrase carries no information on its own, which is why the useful move is always to ask what the thing does when nobody else participates.
Which of the four meanings involves a vendor?
The first two almost always do, and the third sometimes does when someone sells a hosted execution product. The fourth has no vendor by construction, because there is no service being delivered. If somebody presents themselves as a vendor in the fourth case, what they are selling is access to a group rather than a deliverable.
Do all four leave a trail on the blockchain?
All four leave trades, because every swap is a public transaction. What differs is how much the trail tells you. Tooling leaves a trail that a vendor can point at and say this is the job you paid for. An arrangement leaves a trail that shows a burst of buying and then selling, but does not label who was standing where or who knew what.
Is one of the four safe and the others dangerous?
That framing is too clean. The tooling categories carry ordinary commercial and custody risks that vary enormously by vendor, and any of them can be run badly. The fourth category is different in kind rather than in degree, because its structure makes most participants losers regardless of how well it is executed.
Why does the same person use the word for two different things?
Because the phrase entered general use through conversation rather than documentation, and conversation does not enforce definitions. Most people using it loosely are not trying to mislead anyone. The problem is that a loose word is also convenient for anyone who is, and you cannot tell the two apart by tone.
If I only remember one distinction, which should it be?
Ask whether the thing has customers or participants. A category with customers has someone buying a described job. A category with participants has people contributing their own market orders in exchange for a position in a sequence. Everything else follows from that one difference.
Written by The Pump Bot Primer Desk. Category descriptions here are based on what these products publish about themselves and on transaction behaviour that anyone can verify in public block data. No named company, group or person is characterised as an offender anywhere on this page. Definitions are collected in the term list, and the editorial limits of this desk are described in this desk.